1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * fs/f2fs/file.c 4 * 5 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 6 * http://www.samsung.com/ 7 */ 8 #include <linux/fs.h> 9 #include <linux/f2fs_fs.h> 10 #include <linux/stat.h> 11 #include <linux/buffer_head.h> 12 #include <linux/writeback.h> 13 #include <linux/blkdev.h> 14 #include <linux/falloc.h> 15 #include <linux/types.h> 16 #include <linux/compat.h> 17 #include <linux/uaccess.h> 18 #include <linux/mount.h> 19 #include <linux/pagevec.h> 20 #include <linux/uio.h> 21 #include <linux/uuid.h> 22 #include <linux/file.h> 23 #include <linux/nls.h> 24 #include <linux/sched/signal.h> 25 #include <linux/fileattr.h> 26 #include <linux/fadvise.h> 27 #include <linux/iomap.h> 28 29 #include "f2fs.h" 30 #include "node.h" 31 #include "segment.h" 32 #include "xattr.h" 33 #include "acl.h" 34 #include "gc.h" 35 #include "iostat.h" 36 #include <trace/events/f2fs.h> 37 #include <uapi/linux/f2fs.h> 38 39 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf) 40 { 41 struct inode *inode = file_inode(vmf->vma->vm_file); 42 vm_flags_t flags = vmf->vma->vm_flags; 43 vm_fault_t ret; 44 45 ret = filemap_fault(vmf); 46 if (ret & VM_FAULT_LOCKED) 47 f2fs_update_iostat(F2FS_I_SB(inode), inode, 48 APP_MAPPED_READ_IO, F2FS_BLKSIZE); 49 50 trace_f2fs_filemap_fault(inode, vmf->pgoff, flags, ret); 51 52 return ret; 53 } 54 55 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf) 56 { 57 struct page *page = vmf->page; 58 struct inode *inode = file_inode(vmf->vma->vm_file); 59 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 60 struct dnode_of_data dn; 61 bool need_alloc = !f2fs_is_pinned_file(inode); 62 int err = 0; 63 vm_fault_t ret; 64 65 if (unlikely(IS_IMMUTABLE(inode))) 66 return VM_FAULT_SIGBUS; 67 68 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 69 err = -EIO; 70 goto out; 71 } 72 73 if (unlikely(f2fs_cp_error(sbi))) { 74 err = -EIO; 75 goto out; 76 } 77 78 if (!f2fs_is_checkpoint_ready(sbi)) { 79 err = -ENOSPC; 80 goto out; 81 } 82 83 err = f2fs_convert_inline_inode(inode); 84 if (err) 85 goto out; 86 87 #ifdef CONFIG_F2FS_FS_COMPRESSION 88 if (f2fs_compressed_file(inode)) { 89 int ret = f2fs_is_compressed_cluster(inode, page->index); 90 91 if (ret < 0) { 92 err = ret; 93 goto out; 94 } else if (ret) { 95 need_alloc = false; 96 } 97 } 98 #endif 99 /* should do out of any locked page */ 100 if (need_alloc) 101 f2fs_balance_fs(sbi, true); 102 103 sb_start_pagefault(inode->i_sb); 104 105 f2fs_bug_on(sbi, f2fs_has_inline_data(inode)); 106 107 file_update_time(vmf->vma->vm_file); 108 filemap_invalidate_lock_shared(inode->i_mapping); 109 lock_page(page); 110 if (unlikely(page->mapping != inode->i_mapping || 111 page_offset(page) > i_size_read(inode) || 112 !PageUptodate(page))) { 113 unlock_page(page); 114 err = -EFAULT; 115 goto out_sem; 116 } 117 118 set_new_dnode(&dn, inode, NULL, NULL, 0); 119 if (need_alloc) { 120 /* block allocation */ 121 err = f2fs_get_block_locked(&dn, page->index); 122 } else { 123 err = f2fs_get_dnode_of_data(&dn, page->index, LOOKUP_NODE); 124 f2fs_put_dnode(&dn); 125 if (f2fs_is_pinned_file(inode) && 126 !__is_valid_data_blkaddr(dn.data_blkaddr)) 127 err = -EIO; 128 } 129 130 if (err) { 131 unlock_page(page); 132 goto out_sem; 133 } 134 135 f2fs_wait_on_page_writeback(page, DATA, false, true); 136 137 /* wait for GCed page writeback via META_MAPPING */ 138 f2fs_wait_on_block_writeback(inode, dn.data_blkaddr); 139 140 /* 141 * check to see if the page is mapped already (no holes) 142 */ 143 if (PageMappedToDisk(page)) 144 goto out_sem; 145 146 /* page is wholly or partially inside EOF */ 147 if (((loff_t)(page->index + 1) << PAGE_SHIFT) > 148 i_size_read(inode)) { 149 loff_t offset; 150 151 offset = i_size_read(inode) & ~PAGE_MASK; 152 zero_user_segment(page, offset, PAGE_SIZE); 153 } 154 set_page_dirty(page); 155 156 f2fs_update_iostat(sbi, inode, APP_MAPPED_IO, F2FS_BLKSIZE); 157 f2fs_update_time(sbi, REQ_TIME); 158 159 out_sem: 160 filemap_invalidate_unlock_shared(inode->i_mapping); 161 162 sb_end_pagefault(inode->i_sb); 163 out: 164 ret = vmf_fs_error(err); 165 166 trace_f2fs_vm_page_mkwrite(inode, page->index, vmf->vma->vm_flags, ret); 167 return ret; 168 } 169 170 static const struct vm_operations_struct f2fs_file_vm_ops = { 171 .fault = f2fs_filemap_fault, 172 .map_pages = filemap_map_pages, 173 .page_mkwrite = f2fs_vm_page_mkwrite, 174 }; 175 176 static int get_parent_ino(struct inode *inode, nid_t *pino) 177 { 178 struct dentry *dentry; 179 180 /* 181 * Make sure to get the non-deleted alias. The alias associated with 182 * the open file descriptor being fsync()'ed may be deleted already. 183 */ 184 dentry = d_find_alias(inode); 185 if (!dentry) 186 return 0; 187 188 *pino = d_parent_ino(dentry); 189 dput(dentry); 190 return 1; 191 } 192 193 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode) 194 { 195 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 196 enum cp_reason_type cp_reason = CP_NO_NEEDED; 197 198 if (!S_ISREG(inode->i_mode)) 199 cp_reason = CP_NON_REGULAR; 200 else if (f2fs_compressed_file(inode)) 201 cp_reason = CP_COMPRESSED; 202 else if (inode->i_nlink != 1) 203 cp_reason = CP_HARDLINK; 204 else if (is_sbi_flag_set(sbi, SBI_NEED_CP)) 205 cp_reason = CP_SB_NEED_CP; 206 else if (file_wrong_pino(inode)) 207 cp_reason = CP_WRONG_PINO; 208 else if (!f2fs_space_for_roll_forward(sbi)) 209 cp_reason = CP_NO_SPC_ROLL; 210 else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino)) 211 cp_reason = CP_NODE_NEED_CP; 212 else if (test_opt(sbi, FASTBOOT)) 213 cp_reason = CP_FASTBOOT_MODE; 214 else if (F2FS_OPTION(sbi).active_logs == 2) 215 cp_reason = CP_SPEC_LOG_NUM; 216 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT && 217 f2fs_need_dentry_mark(sbi, inode->i_ino) && 218 f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino, 219 TRANS_DIR_INO)) 220 cp_reason = CP_RECOVER_DIR; 221 222 return cp_reason; 223 } 224 225 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino) 226 { 227 struct page *i = find_get_page(NODE_MAPPING(sbi), ino); 228 bool ret = false; 229 /* But we need to avoid that there are some inode updates */ 230 if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino)) 231 ret = true; 232 f2fs_put_page(i, 0); 233 return ret; 234 } 235 236 static void try_to_fix_pino(struct inode *inode) 237 { 238 struct f2fs_inode_info *fi = F2FS_I(inode); 239 nid_t pino; 240 241 f2fs_down_write(&fi->i_sem); 242 if (file_wrong_pino(inode) && inode->i_nlink == 1 && 243 get_parent_ino(inode, &pino)) { 244 f2fs_i_pino_write(inode, pino); 245 file_got_pino(inode); 246 } 247 f2fs_up_write(&fi->i_sem); 248 } 249 250 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, 251 int datasync, bool atomic) 252 { 253 struct inode *inode = file->f_mapping->host; 254 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 255 nid_t ino = inode->i_ino; 256 int ret = 0; 257 enum cp_reason_type cp_reason = 0; 258 struct writeback_control wbc = { 259 .sync_mode = WB_SYNC_ALL, 260 .nr_to_write = LONG_MAX, 261 .for_reclaim = 0, 262 }; 263 unsigned int seq_id = 0; 264 265 if (unlikely(f2fs_readonly(inode->i_sb))) 266 return 0; 267 268 trace_f2fs_sync_file_enter(inode); 269 270 if (S_ISDIR(inode->i_mode)) 271 goto go_write; 272 273 /* if fdatasync is triggered, let's do in-place-update */ 274 if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) 275 set_inode_flag(inode, FI_NEED_IPU); 276 ret = file_write_and_wait_range(file, start, end); 277 clear_inode_flag(inode, FI_NEED_IPU); 278 279 if (ret || is_sbi_flag_set(sbi, SBI_CP_DISABLED)) { 280 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 281 return ret; 282 } 283 284 /* if the inode is dirty, let's recover all the time */ 285 if (!f2fs_skip_inode_update(inode, datasync)) { 286 f2fs_write_inode(inode, NULL); 287 goto go_write; 288 } 289 290 /* 291 * if there is no written data, don't waste time to write recovery info. 292 */ 293 if (!is_inode_flag_set(inode, FI_APPEND_WRITE) && 294 !f2fs_exist_written_data(sbi, ino, APPEND_INO)) { 295 296 /* it may call write_inode just prior to fsync */ 297 if (need_inode_page_update(sbi, ino)) 298 goto go_write; 299 300 if (is_inode_flag_set(inode, FI_UPDATE_WRITE) || 301 f2fs_exist_written_data(sbi, ino, UPDATE_INO)) 302 goto flush_out; 303 goto out; 304 } else { 305 /* 306 * for OPU case, during fsync(), node can be persisted before 307 * data when lower device doesn't support write barrier, result 308 * in data corruption after SPO. 309 * So for strict fsync mode, force to use atomic write semantics 310 * to keep write order in between data/node and last node to 311 * avoid potential data corruption. 312 */ 313 if (F2FS_OPTION(sbi).fsync_mode == 314 FSYNC_MODE_STRICT && !atomic) 315 atomic = true; 316 } 317 go_write: 318 /* 319 * Both of fdatasync() and fsync() are able to be recovered from 320 * sudden-power-off. 321 */ 322 f2fs_down_read(&F2FS_I(inode)->i_sem); 323 cp_reason = need_do_checkpoint(inode); 324 f2fs_up_read(&F2FS_I(inode)->i_sem); 325 326 if (cp_reason) { 327 /* all the dirty node pages should be flushed for POR */ 328 ret = f2fs_sync_fs(inode->i_sb, 1); 329 330 /* 331 * We've secured consistency through sync_fs. Following pino 332 * will be used only for fsynced inodes after checkpoint. 333 */ 334 try_to_fix_pino(inode); 335 clear_inode_flag(inode, FI_APPEND_WRITE); 336 clear_inode_flag(inode, FI_UPDATE_WRITE); 337 goto out; 338 } 339 sync_nodes: 340 atomic_inc(&sbi->wb_sync_req[NODE]); 341 ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); 342 atomic_dec(&sbi->wb_sync_req[NODE]); 343 if (ret) 344 goto out; 345 346 /* if cp_error was enabled, we should avoid infinite loop */ 347 if (unlikely(f2fs_cp_error(sbi))) { 348 ret = -EIO; 349 goto out; 350 } 351 352 if (f2fs_need_inode_block_update(sbi, ino)) { 353 f2fs_mark_inode_dirty_sync(inode, true); 354 f2fs_write_inode(inode, NULL); 355 goto sync_nodes; 356 } 357 358 /* 359 * If it's atomic_write, it's just fine to keep write ordering. So 360 * here we don't need to wait for node write completion, since we use 361 * node chain which serializes node blocks. If one of node writes are 362 * reordered, we can see simply broken chain, resulting in stopping 363 * roll-forward recovery. It means we'll recover all or none node blocks 364 * given fsync mark. 365 */ 366 if (!atomic) { 367 ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); 368 if (ret) 369 goto out; 370 } 371 372 /* once recovery info is written, don't need to tack this */ 373 f2fs_remove_ino_entry(sbi, ino, APPEND_INO); 374 clear_inode_flag(inode, FI_APPEND_WRITE); 375 flush_out: 376 if ((!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER) || 377 (atomic && !test_opt(sbi, NOBARRIER) && f2fs_sb_has_blkzoned(sbi))) 378 ret = f2fs_issue_flush(sbi, inode->i_ino); 379 if (!ret) { 380 f2fs_remove_ino_entry(sbi, ino, UPDATE_INO); 381 clear_inode_flag(inode, FI_UPDATE_WRITE); 382 f2fs_remove_ino_entry(sbi, ino, FLUSH_INO); 383 } 384 f2fs_update_time(sbi, REQ_TIME); 385 out: 386 trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret); 387 return ret; 388 } 389 390 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) 391 { 392 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 393 return -EIO; 394 return f2fs_do_sync_file(file, start, end, datasync, false); 395 } 396 397 static bool __found_offset(struct address_space *mapping, 398 struct dnode_of_data *dn, pgoff_t index, int whence) 399 { 400 block_t blkaddr = f2fs_data_blkaddr(dn); 401 struct inode *inode = mapping->host; 402 bool compressed_cluster = false; 403 404 if (f2fs_compressed_file(inode)) { 405 block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_page, 406 ALIGN_DOWN(dn->ofs_in_node, F2FS_I(inode)->i_cluster_size)); 407 408 compressed_cluster = first_blkaddr == COMPRESS_ADDR; 409 } 410 411 switch (whence) { 412 case SEEK_DATA: 413 if (__is_valid_data_blkaddr(blkaddr)) 414 return true; 415 if (blkaddr == NEW_ADDR && 416 xa_get_mark(&mapping->i_pages, index, PAGECACHE_TAG_DIRTY)) 417 return true; 418 if (compressed_cluster) 419 return true; 420 break; 421 case SEEK_HOLE: 422 if (compressed_cluster) 423 return false; 424 if (blkaddr == NULL_ADDR) 425 return true; 426 break; 427 } 428 return false; 429 } 430 431 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence) 432 { 433 struct inode *inode = file->f_mapping->host; 434 loff_t maxbytes = inode->i_sb->s_maxbytes; 435 struct dnode_of_data dn; 436 pgoff_t pgofs, end_offset; 437 loff_t data_ofs = offset; 438 loff_t isize; 439 int err = 0; 440 441 inode_lock_shared(inode); 442 443 isize = i_size_read(inode); 444 if (offset >= isize) 445 goto fail; 446 447 /* handle inline data case */ 448 if (f2fs_has_inline_data(inode)) { 449 if (whence == SEEK_HOLE) { 450 data_ofs = isize; 451 goto found; 452 } else if (whence == SEEK_DATA) { 453 data_ofs = offset; 454 goto found; 455 } 456 } 457 458 pgofs = (pgoff_t)(offset >> PAGE_SHIFT); 459 460 for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 461 set_new_dnode(&dn, inode, NULL, NULL, 0); 462 err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE); 463 if (err && err != -ENOENT) { 464 goto fail; 465 } else if (err == -ENOENT) { 466 /* direct node does not exists */ 467 if (whence == SEEK_DATA) { 468 pgofs = f2fs_get_next_page_offset(&dn, pgofs); 469 continue; 470 } else { 471 goto found; 472 } 473 } 474 475 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 476 477 /* find data/hole in dnode block */ 478 for (; dn.ofs_in_node < end_offset; 479 dn.ofs_in_node++, pgofs++, 480 data_ofs = (loff_t)pgofs << PAGE_SHIFT) { 481 block_t blkaddr; 482 483 blkaddr = f2fs_data_blkaddr(&dn); 484 485 if (__is_valid_data_blkaddr(blkaddr) && 486 !f2fs_is_valid_blkaddr(F2FS_I_SB(inode), 487 blkaddr, DATA_GENERIC_ENHANCE)) { 488 f2fs_put_dnode(&dn); 489 goto fail; 490 } 491 492 if (__found_offset(file->f_mapping, &dn, 493 pgofs, whence)) { 494 f2fs_put_dnode(&dn); 495 goto found; 496 } 497 } 498 f2fs_put_dnode(&dn); 499 } 500 501 if (whence == SEEK_DATA) 502 goto fail; 503 found: 504 if (whence == SEEK_HOLE && data_ofs > isize) 505 data_ofs = isize; 506 inode_unlock_shared(inode); 507 return vfs_setpos(file, data_ofs, maxbytes); 508 fail: 509 inode_unlock_shared(inode); 510 return -ENXIO; 511 } 512 513 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence) 514 { 515 struct inode *inode = file->f_mapping->host; 516 loff_t maxbytes = inode->i_sb->s_maxbytes; 517 518 if (f2fs_compressed_file(inode)) 519 maxbytes = max_file_blocks(inode) << F2FS_BLKSIZE_BITS; 520 521 switch (whence) { 522 case SEEK_SET: 523 case SEEK_CUR: 524 case SEEK_END: 525 return generic_file_llseek_size(file, offset, whence, 526 maxbytes, i_size_read(inode)); 527 case SEEK_DATA: 528 case SEEK_HOLE: 529 if (offset < 0) 530 return -ENXIO; 531 return f2fs_seek_block(file, offset, whence); 532 } 533 534 return -EINVAL; 535 } 536 537 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma) 538 { 539 struct inode *inode = file_inode(file); 540 541 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 542 return -EIO; 543 544 if (!f2fs_is_compress_backend_ready(inode)) 545 return -EOPNOTSUPP; 546 547 file_accessed(file); 548 vma->vm_ops = &f2fs_file_vm_ops; 549 550 f2fs_down_read(&F2FS_I(inode)->i_sem); 551 set_inode_flag(inode, FI_MMAP_FILE); 552 f2fs_up_read(&F2FS_I(inode)->i_sem); 553 554 return 0; 555 } 556 557 static int f2fs_file_open(struct inode *inode, struct file *filp) 558 { 559 int err = fscrypt_file_open(inode, filp); 560 561 if (err) 562 return err; 563 564 if (!f2fs_is_compress_backend_ready(inode)) 565 return -EOPNOTSUPP; 566 567 err = fsverity_file_open(inode, filp); 568 if (err) 569 return err; 570 571 filp->f_mode |= FMODE_NOWAIT; 572 filp->f_mode |= FMODE_CAN_ODIRECT; 573 574 return dquot_file_open(inode, filp); 575 } 576 577 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count) 578 { 579 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 580 int nr_free = 0, ofs = dn->ofs_in_node, len = count; 581 __le32 *addr; 582 bool compressed_cluster = false; 583 int cluster_index = 0, valid_blocks = 0; 584 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 585 bool released = !atomic_read(&F2FS_I(dn->inode)->i_compr_blocks); 586 587 addr = get_dnode_addr(dn->inode, dn->node_page) + ofs; 588 589 /* Assumption: truncation starts with cluster */ 590 for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) { 591 block_t blkaddr = le32_to_cpu(*addr); 592 593 if (f2fs_compressed_file(dn->inode) && 594 !(cluster_index & (cluster_size - 1))) { 595 if (compressed_cluster) 596 f2fs_i_compr_blocks_update(dn->inode, 597 valid_blocks, false); 598 compressed_cluster = (blkaddr == COMPRESS_ADDR); 599 valid_blocks = 0; 600 } 601 602 if (blkaddr == NULL_ADDR) 603 continue; 604 605 f2fs_set_data_blkaddr(dn, NULL_ADDR); 606 607 if (__is_valid_data_blkaddr(blkaddr)) { 608 if (time_to_inject(sbi, FAULT_BLKADDR_CONSISTENCE)) 609 continue; 610 if (!f2fs_is_valid_blkaddr_raw(sbi, blkaddr, 611 DATA_GENERIC_ENHANCE)) 612 continue; 613 if (compressed_cluster) 614 valid_blocks++; 615 } 616 617 f2fs_invalidate_blocks(sbi, blkaddr); 618 619 if (!released || blkaddr != COMPRESS_ADDR) 620 nr_free++; 621 } 622 623 if (compressed_cluster) 624 f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false); 625 626 if (nr_free) { 627 pgoff_t fofs; 628 /* 629 * once we invalidate valid blkaddr in range [ofs, ofs + count], 630 * we will invalidate all blkaddr in the whole range. 631 */ 632 fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), 633 dn->inode) + ofs; 634 f2fs_update_read_extent_cache_range(dn, fofs, 0, len); 635 f2fs_update_age_extent_cache_range(dn, fofs, len); 636 dec_valid_block_count(sbi, dn->inode, nr_free); 637 } 638 dn->ofs_in_node = ofs; 639 640 f2fs_update_time(sbi, REQ_TIME); 641 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid, 642 dn->ofs_in_node, nr_free); 643 } 644 645 static int truncate_partial_data_page(struct inode *inode, u64 from, 646 bool cache_only) 647 { 648 loff_t offset = from & (PAGE_SIZE - 1); 649 pgoff_t index = from >> PAGE_SHIFT; 650 struct address_space *mapping = inode->i_mapping; 651 struct page *page; 652 653 if (!offset && !cache_only) 654 return 0; 655 656 if (cache_only) { 657 page = find_lock_page(mapping, index); 658 if (page && PageUptodate(page)) 659 goto truncate_out; 660 f2fs_put_page(page, 1); 661 return 0; 662 } 663 664 page = f2fs_get_lock_data_page(inode, index, true); 665 if (IS_ERR(page)) 666 return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page); 667 truncate_out: 668 f2fs_wait_on_page_writeback(page, DATA, true, true); 669 zero_user(page, offset, PAGE_SIZE - offset); 670 671 /* An encrypted inode should have a key and truncate the last page. */ 672 f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode)); 673 if (!cache_only) 674 set_page_dirty(page); 675 f2fs_put_page(page, 1); 676 return 0; 677 } 678 679 int f2fs_do_truncate_blocks(struct inode *inode, u64 from, bool lock) 680 { 681 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 682 struct dnode_of_data dn; 683 pgoff_t free_from; 684 int count = 0, err = 0; 685 struct page *ipage; 686 bool truncate_page = false; 687 688 trace_f2fs_truncate_blocks_enter(inode, from); 689 690 free_from = (pgoff_t)F2FS_BLK_ALIGN(from); 691 692 if (free_from >= max_file_blocks(inode)) 693 goto free_partial; 694 695 if (lock) 696 f2fs_lock_op(sbi); 697 698 ipage = f2fs_get_node_page(sbi, inode->i_ino); 699 if (IS_ERR(ipage)) { 700 err = PTR_ERR(ipage); 701 goto out; 702 } 703 704 if (f2fs_has_inline_data(inode)) { 705 f2fs_truncate_inline_inode(inode, ipage, from); 706 f2fs_put_page(ipage, 1); 707 truncate_page = true; 708 goto out; 709 } 710 711 set_new_dnode(&dn, inode, ipage, NULL, 0); 712 err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA); 713 if (err) { 714 if (err == -ENOENT) 715 goto free_next; 716 goto out; 717 } 718 719 count = ADDRS_PER_PAGE(dn.node_page, inode); 720 721 count -= dn.ofs_in_node; 722 f2fs_bug_on(sbi, count < 0); 723 724 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 725 f2fs_truncate_data_blocks_range(&dn, count); 726 free_from += count; 727 } 728 729 f2fs_put_dnode(&dn); 730 free_next: 731 err = f2fs_truncate_inode_blocks(inode, free_from); 732 out: 733 if (lock) 734 f2fs_unlock_op(sbi); 735 free_partial: 736 /* lastly zero out the first data page */ 737 if (!err) 738 err = truncate_partial_data_page(inode, from, truncate_page); 739 740 trace_f2fs_truncate_blocks_exit(inode, err); 741 return err; 742 } 743 744 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock) 745 { 746 u64 free_from = from; 747 int err; 748 749 #ifdef CONFIG_F2FS_FS_COMPRESSION 750 /* 751 * for compressed file, only support cluster size 752 * aligned truncation. 753 */ 754 if (f2fs_compressed_file(inode)) 755 free_from = round_up(from, 756 F2FS_I(inode)->i_cluster_size << PAGE_SHIFT); 757 #endif 758 759 err = f2fs_do_truncate_blocks(inode, free_from, lock); 760 if (err) 761 return err; 762 763 #ifdef CONFIG_F2FS_FS_COMPRESSION 764 /* 765 * For compressed file, after release compress blocks, don't allow write 766 * direct, but we should allow write direct after truncate to zero. 767 */ 768 if (f2fs_compressed_file(inode) && !free_from 769 && is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 770 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 771 772 if (from != free_from) { 773 err = f2fs_truncate_partial_cluster(inode, from, lock); 774 if (err) 775 return err; 776 } 777 #endif 778 779 return 0; 780 } 781 782 int f2fs_truncate(struct inode *inode) 783 { 784 int err; 785 786 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 787 return -EIO; 788 789 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 790 S_ISLNK(inode->i_mode))) 791 return 0; 792 793 trace_f2fs_truncate(inode); 794 795 if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) 796 return -EIO; 797 798 err = f2fs_dquot_initialize(inode); 799 if (err) 800 return err; 801 802 /* we should check inline_data size */ 803 if (!f2fs_may_inline_data(inode)) { 804 err = f2fs_convert_inline_inode(inode); 805 if (err) 806 return err; 807 } 808 809 err = f2fs_truncate_blocks(inode, i_size_read(inode), true); 810 if (err) 811 return err; 812 813 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 814 f2fs_mark_inode_dirty_sync(inode, false); 815 return 0; 816 } 817 818 static bool f2fs_force_buffered_io(struct inode *inode, int rw) 819 { 820 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 821 822 if (!fscrypt_dio_supported(inode)) 823 return true; 824 if (fsverity_active(inode)) 825 return true; 826 if (f2fs_compressed_file(inode)) 827 return true; 828 829 /* disallow direct IO if any of devices has unaligned blksize */ 830 if (f2fs_is_multi_device(sbi) && !sbi->aligned_blksize) 831 return true; 832 /* 833 * for blkzoned device, fallback direct IO to buffered IO, so 834 * all IOs can be serialized by log-structured write. 835 */ 836 if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE) && 837 !f2fs_is_pinned_file(inode)) 838 return true; 839 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED)) 840 return true; 841 842 return false; 843 } 844 845 int f2fs_getattr(struct mnt_idmap *idmap, const struct path *path, 846 struct kstat *stat, u32 request_mask, unsigned int query_flags) 847 { 848 struct inode *inode = d_inode(path->dentry); 849 struct f2fs_inode_info *fi = F2FS_I(inode); 850 struct f2fs_inode *ri = NULL; 851 unsigned int flags; 852 853 if (f2fs_has_extra_attr(inode) && 854 f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) && 855 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) { 856 stat->result_mask |= STATX_BTIME; 857 stat->btime.tv_sec = fi->i_crtime.tv_sec; 858 stat->btime.tv_nsec = fi->i_crtime.tv_nsec; 859 } 860 861 /* 862 * Return the DIO alignment restrictions if requested. We only return 863 * this information when requested, since on encrypted files it might 864 * take a fair bit of work to get if the file wasn't opened recently. 865 * 866 * f2fs sometimes supports DIO reads but not DIO writes. STATX_DIOALIGN 867 * cannot represent that, so in that case we report no DIO support. 868 */ 869 if ((request_mask & STATX_DIOALIGN) && S_ISREG(inode->i_mode)) { 870 unsigned int bsize = i_blocksize(inode); 871 872 stat->result_mask |= STATX_DIOALIGN; 873 if (!f2fs_force_buffered_io(inode, WRITE)) { 874 stat->dio_mem_align = bsize; 875 stat->dio_offset_align = bsize; 876 } 877 } 878 879 flags = fi->i_flags; 880 if (flags & F2FS_COMPR_FL) 881 stat->attributes |= STATX_ATTR_COMPRESSED; 882 if (flags & F2FS_APPEND_FL) 883 stat->attributes |= STATX_ATTR_APPEND; 884 if (IS_ENCRYPTED(inode)) 885 stat->attributes |= STATX_ATTR_ENCRYPTED; 886 if (flags & F2FS_IMMUTABLE_FL) 887 stat->attributes |= STATX_ATTR_IMMUTABLE; 888 if (flags & F2FS_NODUMP_FL) 889 stat->attributes |= STATX_ATTR_NODUMP; 890 if (IS_VERITY(inode)) 891 stat->attributes |= STATX_ATTR_VERITY; 892 893 stat->attributes_mask |= (STATX_ATTR_COMPRESSED | 894 STATX_ATTR_APPEND | 895 STATX_ATTR_ENCRYPTED | 896 STATX_ATTR_IMMUTABLE | 897 STATX_ATTR_NODUMP | 898 STATX_ATTR_VERITY); 899 900 generic_fillattr(idmap, request_mask, inode, stat); 901 902 /* we need to show initial sectors used for inline_data/dentries */ 903 if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) || 904 f2fs_has_inline_dentry(inode)) 905 stat->blocks += (stat->size + 511) >> 9; 906 907 return 0; 908 } 909 910 #ifdef CONFIG_F2FS_FS_POSIX_ACL 911 static void __setattr_copy(struct mnt_idmap *idmap, 912 struct inode *inode, const struct iattr *attr) 913 { 914 unsigned int ia_valid = attr->ia_valid; 915 916 i_uid_update(idmap, attr, inode); 917 i_gid_update(idmap, attr, inode); 918 if (ia_valid & ATTR_ATIME) 919 inode_set_atime_to_ts(inode, attr->ia_atime); 920 if (ia_valid & ATTR_MTIME) 921 inode_set_mtime_to_ts(inode, attr->ia_mtime); 922 if (ia_valid & ATTR_CTIME) 923 inode_set_ctime_to_ts(inode, attr->ia_ctime); 924 if (ia_valid & ATTR_MODE) { 925 umode_t mode = attr->ia_mode; 926 927 if (!in_group_or_capable(idmap, inode, i_gid_into_vfsgid(idmap, inode))) 928 mode &= ~S_ISGID; 929 set_acl_inode(inode, mode); 930 } 931 } 932 #else 933 #define __setattr_copy setattr_copy 934 #endif 935 936 int f2fs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, 937 struct iattr *attr) 938 { 939 struct inode *inode = d_inode(dentry); 940 int err; 941 942 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 943 return -EIO; 944 945 if (unlikely(IS_IMMUTABLE(inode))) 946 return -EPERM; 947 948 if (unlikely(IS_APPEND(inode) && 949 (attr->ia_valid & (ATTR_MODE | ATTR_UID | 950 ATTR_GID | ATTR_TIMES_SET)))) 951 return -EPERM; 952 953 if ((attr->ia_valid & ATTR_SIZE)) { 954 if (!f2fs_is_compress_backend_ready(inode)) 955 return -EOPNOTSUPP; 956 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) && 957 !IS_ALIGNED(attr->ia_size, 958 F2FS_BLK_TO_BYTES(F2FS_I(inode)->i_cluster_size))) 959 return -EINVAL; 960 } 961 962 err = setattr_prepare(idmap, dentry, attr); 963 if (err) 964 return err; 965 966 err = fscrypt_prepare_setattr(dentry, attr); 967 if (err) 968 return err; 969 970 err = fsverity_prepare_setattr(dentry, attr); 971 if (err) 972 return err; 973 974 if (is_quota_modification(idmap, inode, attr)) { 975 err = f2fs_dquot_initialize(inode); 976 if (err) 977 return err; 978 } 979 if (i_uid_needs_update(idmap, attr, inode) || 980 i_gid_needs_update(idmap, attr, inode)) { 981 f2fs_lock_op(F2FS_I_SB(inode)); 982 err = dquot_transfer(idmap, inode, attr); 983 if (err) { 984 set_sbi_flag(F2FS_I_SB(inode), 985 SBI_QUOTA_NEED_REPAIR); 986 f2fs_unlock_op(F2FS_I_SB(inode)); 987 return err; 988 } 989 /* 990 * update uid/gid under lock_op(), so that dquot and inode can 991 * be updated atomically. 992 */ 993 i_uid_update(idmap, attr, inode); 994 i_gid_update(idmap, attr, inode); 995 f2fs_mark_inode_dirty_sync(inode, true); 996 f2fs_unlock_op(F2FS_I_SB(inode)); 997 } 998 999 if (attr->ia_valid & ATTR_SIZE) { 1000 loff_t old_size = i_size_read(inode); 1001 1002 if (attr->ia_size > MAX_INLINE_DATA(inode)) { 1003 /* 1004 * should convert inline inode before i_size_write to 1005 * keep smaller than inline_data size with inline flag. 1006 */ 1007 err = f2fs_convert_inline_inode(inode); 1008 if (err) 1009 return err; 1010 } 1011 1012 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1013 filemap_invalidate_lock(inode->i_mapping); 1014 1015 truncate_setsize(inode, attr->ia_size); 1016 1017 if (attr->ia_size <= old_size) 1018 err = f2fs_truncate(inode); 1019 /* 1020 * do not trim all blocks after i_size if target size is 1021 * larger than i_size. 1022 */ 1023 filemap_invalidate_unlock(inode->i_mapping); 1024 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1025 if (err) 1026 return err; 1027 1028 spin_lock(&F2FS_I(inode)->i_size_lock); 1029 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1030 F2FS_I(inode)->last_disk_size = i_size_read(inode); 1031 spin_unlock(&F2FS_I(inode)->i_size_lock); 1032 } 1033 1034 __setattr_copy(idmap, inode, attr); 1035 1036 if (attr->ia_valid & ATTR_MODE) { 1037 err = posix_acl_chmod(idmap, dentry, f2fs_get_inode_mode(inode)); 1038 1039 if (is_inode_flag_set(inode, FI_ACL_MODE)) { 1040 if (!err) 1041 inode->i_mode = F2FS_I(inode)->i_acl_mode; 1042 clear_inode_flag(inode, FI_ACL_MODE); 1043 } 1044 } 1045 1046 /* file size may changed here */ 1047 f2fs_mark_inode_dirty_sync(inode, true); 1048 1049 /* inode change will produce dirty node pages flushed by checkpoint */ 1050 f2fs_balance_fs(F2FS_I_SB(inode), true); 1051 1052 return err; 1053 } 1054 1055 const struct inode_operations f2fs_file_inode_operations = { 1056 .getattr = f2fs_getattr, 1057 .setattr = f2fs_setattr, 1058 .get_inode_acl = f2fs_get_acl, 1059 .set_acl = f2fs_set_acl, 1060 .listxattr = f2fs_listxattr, 1061 .fiemap = f2fs_fiemap, 1062 .fileattr_get = f2fs_fileattr_get, 1063 .fileattr_set = f2fs_fileattr_set, 1064 }; 1065 1066 static int fill_zero(struct inode *inode, pgoff_t index, 1067 loff_t start, loff_t len) 1068 { 1069 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1070 struct page *page; 1071 1072 if (!len) 1073 return 0; 1074 1075 f2fs_balance_fs(sbi, true); 1076 1077 f2fs_lock_op(sbi); 1078 page = f2fs_get_new_data_page(inode, NULL, index, false); 1079 f2fs_unlock_op(sbi); 1080 1081 if (IS_ERR(page)) 1082 return PTR_ERR(page); 1083 1084 f2fs_wait_on_page_writeback(page, DATA, true, true); 1085 zero_user(page, start, len); 1086 set_page_dirty(page); 1087 f2fs_put_page(page, 1); 1088 return 0; 1089 } 1090 1091 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end) 1092 { 1093 int err; 1094 1095 while (pg_start < pg_end) { 1096 struct dnode_of_data dn; 1097 pgoff_t end_offset, count; 1098 1099 set_new_dnode(&dn, inode, NULL, NULL, 0); 1100 err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE); 1101 if (err) { 1102 if (err == -ENOENT) { 1103 pg_start = f2fs_get_next_page_offset(&dn, 1104 pg_start); 1105 continue; 1106 } 1107 return err; 1108 } 1109 1110 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1111 count = min(end_offset - dn.ofs_in_node, pg_end - pg_start); 1112 1113 f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset); 1114 1115 f2fs_truncate_data_blocks_range(&dn, count); 1116 f2fs_put_dnode(&dn); 1117 1118 pg_start += count; 1119 } 1120 return 0; 1121 } 1122 1123 static int f2fs_punch_hole(struct inode *inode, loff_t offset, loff_t len) 1124 { 1125 pgoff_t pg_start, pg_end; 1126 loff_t off_start, off_end; 1127 int ret; 1128 1129 ret = f2fs_convert_inline_inode(inode); 1130 if (ret) 1131 return ret; 1132 1133 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1134 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1135 1136 off_start = offset & (PAGE_SIZE - 1); 1137 off_end = (offset + len) & (PAGE_SIZE - 1); 1138 1139 if (pg_start == pg_end) { 1140 ret = fill_zero(inode, pg_start, off_start, 1141 off_end - off_start); 1142 if (ret) 1143 return ret; 1144 } else { 1145 if (off_start) { 1146 ret = fill_zero(inode, pg_start++, off_start, 1147 PAGE_SIZE - off_start); 1148 if (ret) 1149 return ret; 1150 } 1151 if (off_end) { 1152 ret = fill_zero(inode, pg_end, 0, off_end); 1153 if (ret) 1154 return ret; 1155 } 1156 1157 if (pg_start < pg_end) { 1158 loff_t blk_start, blk_end; 1159 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1160 1161 f2fs_balance_fs(sbi, true); 1162 1163 blk_start = (loff_t)pg_start << PAGE_SHIFT; 1164 blk_end = (loff_t)pg_end << PAGE_SHIFT; 1165 1166 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1167 filemap_invalidate_lock(inode->i_mapping); 1168 1169 truncate_pagecache_range(inode, blk_start, blk_end - 1); 1170 1171 f2fs_lock_op(sbi); 1172 ret = f2fs_truncate_hole(inode, pg_start, pg_end); 1173 f2fs_unlock_op(sbi); 1174 1175 filemap_invalidate_unlock(inode->i_mapping); 1176 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1177 } 1178 } 1179 1180 return ret; 1181 } 1182 1183 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr, 1184 int *do_replace, pgoff_t off, pgoff_t len) 1185 { 1186 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1187 struct dnode_of_data dn; 1188 int ret, done, i; 1189 1190 next_dnode: 1191 set_new_dnode(&dn, inode, NULL, NULL, 0); 1192 ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA); 1193 if (ret && ret != -ENOENT) { 1194 return ret; 1195 } else if (ret == -ENOENT) { 1196 if (dn.max_level == 0) 1197 return -ENOENT; 1198 done = min((pgoff_t)ADDRS_PER_BLOCK(inode) - 1199 dn.ofs_in_node, len); 1200 blkaddr += done; 1201 do_replace += done; 1202 goto next; 1203 } 1204 1205 done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) - 1206 dn.ofs_in_node, len); 1207 for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) { 1208 *blkaddr = f2fs_data_blkaddr(&dn); 1209 1210 if (__is_valid_data_blkaddr(*blkaddr) && 1211 !f2fs_is_valid_blkaddr(sbi, *blkaddr, 1212 DATA_GENERIC_ENHANCE)) { 1213 f2fs_put_dnode(&dn); 1214 return -EFSCORRUPTED; 1215 } 1216 1217 if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) { 1218 1219 if (f2fs_lfs_mode(sbi)) { 1220 f2fs_put_dnode(&dn); 1221 return -EOPNOTSUPP; 1222 } 1223 1224 /* do not invalidate this block address */ 1225 f2fs_update_data_blkaddr(&dn, NULL_ADDR); 1226 *do_replace = 1; 1227 } 1228 } 1229 f2fs_put_dnode(&dn); 1230 next: 1231 len -= done; 1232 off += done; 1233 if (len) 1234 goto next_dnode; 1235 return 0; 1236 } 1237 1238 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr, 1239 int *do_replace, pgoff_t off, int len) 1240 { 1241 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1242 struct dnode_of_data dn; 1243 int ret, i; 1244 1245 for (i = 0; i < len; i++, do_replace++, blkaddr++) { 1246 if (*do_replace == 0) 1247 continue; 1248 1249 set_new_dnode(&dn, inode, NULL, NULL, 0); 1250 ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA); 1251 if (ret) { 1252 dec_valid_block_count(sbi, inode, 1); 1253 f2fs_invalidate_blocks(sbi, *blkaddr); 1254 } else { 1255 f2fs_update_data_blkaddr(&dn, *blkaddr); 1256 } 1257 f2fs_put_dnode(&dn); 1258 } 1259 return 0; 1260 } 1261 1262 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode, 1263 block_t *blkaddr, int *do_replace, 1264 pgoff_t src, pgoff_t dst, pgoff_t len, bool full) 1265 { 1266 struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode); 1267 pgoff_t i = 0; 1268 int ret; 1269 1270 while (i < len) { 1271 if (blkaddr[i] == NULL_ADDR && !full) { 1272 i++; 1273 continue; 1274 } 1275 1276 if (do_replace[i] || blkaddr[i] == NULL_ADDR) { 1277 struct dnode_of_data dn; 1278 struct node_info ni; 1279 size_t new_size; 1280 pgoff_t ilen; 1281 1282 set_new_dnode(&dn, dst_inode, NULL, NULL, 0); 1283 ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE); 1284 if (ret) 1285 return ret; 1286 1287 ret = f2fs_get_node_info(sbi, dn.nid, &ni, false); 1288 if (ret) { 1289 f2fs_put_dnode(&dn); 1290 return ret; 1291 } 1292 1293 ilen = min((pgoff_t) 1294 ADDRS_PER_PAGE(dn.node_page, dst_inode) - 1295 dn.ofs_in_node, len - i); 1296 do { 1297 dn.data_blkaddr = f2fs_data_blkaddr(&dn); 1298 f2fs_truncate_data_blocks_range(&dn, 1); 1299 1300 if (do_replace[i]) { 1301 f2fs_i_blocks_write(src_inode, 1302 1, false, false); 1303 f2fs_i_blocks_write(dst_inode, 1304 1, true, false); 1305 f2fs_replace_block(sbi, &dn, dn.data_blkaddr, 1306 blkaddr[i], ni.version, true, false); 1307 1308 do_replace[i] = 0; 1309 } 1310 dn.ofs_in_node++; 1311 i++; 1312 new_size = (loff_t)(dst + i) << PAGE_SHIFT; 1313 if (dst_inode->i_size < new_size) 1314 f2fs_i_size_write(dst_inode, new_size); 1315 } while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR)); 1316 1317 f2fs_put_dnode(&dn); 1318 } else { 1319 struct page *psrc, *pdst; 1320 1321 psrc = f2fs_get_lock_data_page(src_inode, 1322 src + i, true); 1323 if (IS_ERR(psrc)) 1324 return PTR_ERR(psrc); 1325 pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i, 1326 true); 1327 if (IS_ERR(pdst)) { 1328 f2fs_put_page(psrc, 1); 1329 return PTR_ERR(pdst); 1330 } 1331 1332 f2fs_wait_on_page_writeback(pdst, DATA, true, true); 1333 1334 memcpy_page(pdst, 0, psrc, 0, PAGE_SIZE); 1335 set_page_dirty(pdst); 1336 set_page_private_gcing(pdst); 1337 f2fs_put_page(pdst, 1); 1338 f2fs_put_page(psrc, 1); 1339 1340 ret = f2fs_truncate_hole(src_inode, 1341 src + i, src + i + 1); 1342 if (ret) 1343 return ret; 1344 i++; 1345 } 1346 } 1347 return 0; 1348 } 1349 1350 static int __exchange_data_block(struct inode *src_inode, 1351 struct inode *dst_inode, pgoff_t src, pgoff_t dst, 1352 pgoff_t len, bool full) 1353 { 1354 block_t *src_blkaddr; 1355 int *do_replace; 1356 pgoff_t olen; 1357 int ret; 1358 1359 while (len) { 1360 olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len); 1361 1362 src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1363 array_size(olen, sizeof(block_t)), 1364 GFP_NOFS); 1365 if (!src_blkaddr) 1366 return -ENOMEM; 1367 1368 do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode), 1369 array_size(olen, sizeof(int)), 1370 GFP_NOFS); 1371 if (!do_replace) { 1372 kvfree(src_blkaddr); 1373 return -ENOMEM; 1374 } 1375 1376 ret = __read_out_blkaddrs(src_inode, src_blkaddr, 1377 do_replace, src, olen); 1378 if (ret) 1379 goto roll_back; 1380 1381 ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr, 1382 do_replace, src, dst, olen, full); 1383 if (ret) 1384 goto roll_back; 1385 1386 src += olen; 1387 dst += olen; 1388 len -= olen; 1389 1390 kvfree(src_blkaddr); 1391 kvfree(do_replace); 1392 } 1393 return 0; 1394 1395 roll_back: 1396 __roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen); 1397 kvfree(src_blkaddr); 1398 kvfree(do_replace); 1399 return ret; 1400 } 1401 1402 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len) 1403 { 1404 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1405 pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1406 pgoff_t start = offset >> PAGE_SHIFT; 1407 pgoff_t end = (offset + len) >> PAGE_SHIFT; 1408 int ret; 1409 1410 f2fs_balance_fs(sbi, true); 1411 1412 /* avoid gc operation during block exchange */ 1413 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1414 filemap_invalidate_lock(inode->i_mapping); 1415 1416 f2fs_lock_op(sbi); 1417 f2fs_drop_extent_tree(inode); 1418 truncate_pagecache(inode, offset); 1419 ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true); 1420 f2fs_unlock_op(sbi); 1421 1422 filemap_invalidate_unlock(inode->i_mapping); 1423 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1424 return ret; 1425 } 1426 1427 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len) 1428 { 1429 loff_t new_size; 1430 int ret; 1431 1432 if (offset + len >= i_size_read(inode)) 1433 return -EINVAL; 1434 1435 /* collapse range should be aligned to block size of f2fs. */ 1436 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1437 return -EINVAL; 1438 1439 ret = f2fs_convert_inline_inode(inode); 1440 if (ret) 1441 return ret; 1442 1443 /* write out all dirty pages from offset */ 1444 ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1445 if (ret) 1446 return ret; 1447 1448 ret = f2fs_do_collapse(inode, offset, len); 1449 if (ret) 1450 return ret; 1451 1452 /* write out all moved pages, if possible */ 1453 filemap_invalidate_lock(inode->i_mapping); 1454 filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX); 1455 truncate_pagecache(inode, offset); 1456 1457 new_size = i_size_read(inode) - len; 1458 ret = f2fs_truncate_blocks(inode, new_size, true); 1459 filemap_invalidate_unlock(inode->i_mapping); 1460 if (!ret) 1461 f2fs_i_size_write(inode, new_size); 1462 return ret; 1463 } 1464 1465 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start, 1466 pgoff_t end) 1467 { 1468 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 1469 pgoff_t index = start; 1470 unsigned int ofs_in_node = dn->ofs_in_node; 1471 blkcnt_t count = 0; 1472 int ret; 1473 1474 for (; index < end; index++, dn->ofs_in_node++) { 1475 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 1476 count++; 1477 } 1478 1479 dn->ofs_in_node = ofs_in_node; 1480 ret = f2fs_reserve_new_blocks(dn, count); 1481 if (ret) 1482 return ret; 1483 1484 dn->ofs_in_node = ofs_in_node; 1485 for (index = start; index < end; index++, dn->ofs_in_node++) { 1486 dn->data_blkaddr = f2fs_data_blkaddr(dn); 1487 /* 1488 * f2fs_reserve_new_blocks will not guarantee entire block 1489 * allocation. 1490 */ 1491 if (dn->data_blkaddr == NULL_ADDR) { 1492 ret = -ENOSPC; 1493 break; 1494 } 1495 1496 if (dn->data_blkaddr == NEW_ADDR) 1497 continue; 1498 1499 if (!f2fs_is_valid_blkaddr(sbi, dn->data_blkaddr, 1500 DATA_GENERIC_ENHANCE)) { 1501 ret = -EFSCORRUPTED; 1502 break; 1503 } 1504 1505 f2fs_invalidate_blocks(sbi, dn->data_blkaddr); 1506 f2fs_set_data_blkaddr(dn, NEW_ADDR); 1507 } 1508 1509 f2fs_update_read_extent_cache_range(dn, start, 0, index - start); 1510 f2fs_update_age_extent_cache_range(dn, start, index - start); 1511 1512 return ret; 1513 } 1514 1515 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len, 1516 int mode) 1517 { 1518 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1519 struct address_space *mapping = inode->i_mapping; 1520 pgoff_t index, pg_start, pg_end; 1521 loff_t new_size = i_size_read(inode); 1522 loff_t off_start, off_end; 1523 int ret = 0; 1524 1525 ret = inode_newsize_ok(inode, (len + offset)); 1526 if (ret) 1527 return ret; 1528 1529 ret = f2fs_convert_inline_inode(inode); 1530 if (ret) 1531 return ret; 1532 1533 ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1); 1534 if (ret) 1535 return ret; 1536 1537 pg_start = ((unsigned long long) offset) >> PAGE_SHIFT; 1538 pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT; 1539 1540 off_start = offset & (PAGE_SIZE - 1); 1541 off_end = (offset + len) & (PAGE_SIZE - 1); 1542 1543 if (pg_start == pg_end) { 1544 ret = fill_zero(inode, pg_start, off_start, 1545 off_end - off_start); 1546 if (ret) 1547 return ret; 1548 1549 new_size = max_t(loff_t, new_size, offset + len); 1550 } else { 1551 if (off_start) { 1552 ret = fill_zero(inode, pg_start++, off_start, 1553 PAGE_SIZE - off_start); 1554 if (ret) 1555 return ret; 1556 1557 new_size = max_t(loff_t, new_size, 1558 (loff_t)pg_start << PAGE_SHIFT); 1559 } 1560 1561 for (index = pg_start; index < pg_end;) { 1562 struct dnode_of_data dn; 1563 unsigned int end_offset; 1564 pgoff_t end; 1565 1566 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1567 filemap_invalidate_lock(mapping); 1568 1569 truncate_pagecache_range(inode, 1570 (loff_t)index << PAGE_SHIFT, 1571 ((loff_t)pg_end << PAGE_SHIFT) - 1); 1572 1573 f2fs_lock_op(sbi); 1574 1575 set_new_dnode(&dn, inode, NULL, NULL, 0); 1576 ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE); 1577 if (ret) { 1578 f2fs_unlock_op(sbi); 1579 filemap_invalidate_unlock(mapping); 1580 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1581 goto out; 1582 } 1583 1584 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 1585 end = min(pg_end, end_offset - dn.ofs_in_node + index); 1586 1587 ret = f2fs_do_zero_range(&dn, index, end); 1588 f2fs_put_dnode(&dn); 1589 1590 f2fs_unlock_op(sbi); 1591 filemap_invalidate_unlock(mapping); 1592 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1593 1594 f2fs_balance_fs(sbi, dn.node_changed); 1595 1596 if (ret) 1597 goto out; 1598 1599 index = end; 1600 new_size = max_t(loff_t, new_size, 1601 (loff_t)index << PAGE_SHIFT); 1602 } 1603 1604 if (off_end) { 1605 ret = fill_zero(inode, pg_end, 0, off_end); 1606 if (ret) 1607 goto out; 1608 1609 new_size = max_t(loff_t, new_size, offset + len); 1610 } 1611 } 1612 1613 out: 1614 if (new_size > i_size_read(inode)) { 1615 if (mode & FALLOC_FL_KEEP_SIZE) 1616 file_set_keep_isize(inode); 1617 else 1618 f2fs_i_size_write(inode, new_size); 1619 } 1620 return ret; 1621 } 1622 1623 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len) 1624 { 1625 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1626 struct address_space *mapping = inode->i_mapping; 1627 pgoff_t nr, pg_start, pg_end, delta, idx; 1628 loff_t new_size; 1629 int ret = 0; 1630 1631 new_size = i_size_read(inode) + len; 1632 ret = inode_newsize_ok(inode, new_size); 1633 if (ret) 1634 return ret; 1635 1636 if (offset >= i_size_read(inode)) 1637 return -EINVAL; 1638 1639 /* insert range should be aligned to block size of f2fs. */ 1640 if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1)) 1641 return -EINVAL; 1642 1643 ret = f2fs_convert_inline_inode(inode); 1644 if (ret) 1645 return ret; 1646 1647 f2fs_balance_fs(sbi, true); 1648 1649 filemap_invalidate_lock(mapping); 1650 ret = f2fs_truncate_blocks(inode, i_size_read(inode), true); 1651 filemap_invalidate_unlock(mapping); 1652 if (ret) 1653 return ret; 1654 1655 /* write out all dirty pages from offset */ 1656 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1657 if (ret) 1658 return ret; 1659 1660 pg_start = offset >> PAGE_SHIFT; 1661 pg_end = (offset + len) >> PAGE_SHIFT; 1662 delta = pg_end - pg_start; 1663 idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 1664 1665 /* avoid gc operation during block exchange */ 1666 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1667 filemap_invalidate_lock(mapping); 1668 truncate_pagecache(inode, offset); 1669 1670 while (!ret && idx > pg_start) { 1671 nr = idx - pg_start; 1672 if (nr > delta) 1673 nr = delta; 1674 idx -= nr; 1675 1676 f2fs_lock_op(sbi); 1677 f2fs_drop_extent_tree(inode); 1678 1679 ret = __exchange_data_block(inode, inode, idx, 1680 idx + delta, nr, false); 1681 f2fs_unlock_op(sbi); 1682 } 1683 filemap_invalidate_unlock(mapping); 1684 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 1685 if (ret) 1686 return ret; 1687 1688 /* write out all moved pages, if possible */ 1689 filemap_invalidate_lock(mapping); 1690 ret = filemap_write_and_wait_range(mapping, offset, LLONG_MAX); 1691 truncate_pagecache(inode, offset); 1692 filemap_invalidate_unlock(mapping); 1693 1694 if (!ret) 1695 f2fs_i_size_write(inode, new_size); 1696 return ret; 1697 } 1698 1699 static int f2fs_expand_inode_data(struct inode *inode, loff_t offset, 1700 loff_t len, int mode) 1701 { 1702 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1703 struct f2fs_map_blocks map = { .m_next_pgofs = NULL, 1704 .m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE, 1705 .m_may_create = true }; 1706 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 1707 .init_gc_type = FG_GC, 1708 .should_migrate_blocks = false, 1709 .err_gc_skipped = true, 1710 .nr_free_secs = 0 }; 1711 pgoff_t pg_start, pg_end; 1712 loff_t new_size; 1713 loff_t off_end; 1714 block_t expanded = 0; 1715 int err; 1716 1717 err = inode_newsize_ok(inode, (len + offset)); 1718 if (err) 1719 return err; 1720 1721 err = f2fs_convert_inline_inode(inode); 1722 if (err) 1723 return err; 1724 1725 f2fs_balance_fs(sbi, true); 1726 1727 pg_start = ((unsigned long long)offset) >> PAGE_SHIFT; 1728 pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT; 1729 off_end = (offset + len) & (PAGE_SIZE - 1); 1730 1731 map.m_lblk = pg_start; 1732 map.m_len = pg_end - pg_start; 1733 if (off_end) 1734 map.m_len++; 1735 1736 if (!map.m_len) 1737 return 0; 1738 1739 if (f2fs_is_pinned_file(inode)) { 1740 block_t sec_blks = CAP_BLKS_PER_SEC(sbi); 1741 block_t sec_len = roundup(map.m_len, sec_blks); 1742 1743 map.m_len = sec_blks; 1744 next_alloc: 1745 if (has_not_enough_free_secs(sbi, 0, 1746 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) { 1747 f2fs_down_write(&sbi->gc_lock); 1748 stat_inc_gc_call_count(sbi, FOREGROUND); 1749 err = f2fs_gc(sbi, &gc_control); 1750 if (err && err != -ENODATA) 1751 goto out_err; 1752 } 1753 1754 f2fs_down_write(&sbi->pin_sem); 1755 1756 err = f2fs_allocate_pinning_section(sbi); 1757 if (err) { 1758 f2fs_up_write(&sbi->pin_sem); 1759 goto out_err; 1760 } 1761 1762 map.m_seg_type = CURSEG_COLD_DATA_PINNED; 1763 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_DIO); 1764 file_dont_truncate(inode); 1765 1766 f2fs_up_write(&sbi->pin_sem); 1767 1768 expanded += map.m_len; 1769 sec_len -= map.m_len; 1770 map.m_lblk += map.m_len; 1771 if (!err && sec_len) 1772 goto next_alloc; 1773 1774 map.m_len = expanded; 1775 } else { 1776 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRE_AIO); 1777 expanded = map.m_len; 1778 } 1779 out_err: 1780 if (err) { 1781 pgoff_t last_off; 1782 1783 if (!expanded) 1784 return err; 1785 1786 last_off = pg_start + expanded - 1; 1787 1788 /* update new size to the failed position */ 1789 new_size = (last_off == pg_end) ? offset + len : 1790 (loff_t)(last_off + 1) << PAGE_SHIFT; 1791 } else { 1792 new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end; 1793 } 1794 1795 if (new_size > i_size_read(inode)) { 1796 if (mode & FALLOC_FL_KEEP_SIZE) 1797 file_set_keep_isize(inode); 1798 else 1799 f2fs_i_size_write(inode, new_size); 1800 } 1801 1802 return err; 1803 } 1804 1805 static long f2fs_fallocate(struct file *file, int mode, 1806 loff_t offset, loff_t len) 1807 { 1808 struct inode *inode = file_inode(file); 1809 long ret = 0; 1810 1811 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 1812 return -EIO; 1813 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 1814 return -ENOSPC; 1815 if (!f2fs_is_compress_backend_ready(inode)) 1816 return -EOPNOTSUPP; 1817 1818 /* f2fs only support ->fallocate for regular file */ 1819 if (!S_ISREG(inode->i_mode)) 1820 return -EINVAL; 1821 1822 if (IS_ENCRYPTED(inode) && 1823 (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE))) 1824 return -EOPNOTSUPP; 1825 1826 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | 1827 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | 1828 FALLOC_FL_INSERT_RANGE)) 1829 return -EOPNOTSUPP; 1830 1831 inode_lock(inode); 1832 1833 /* 1834 * Pinned file should not support partial truncation since the block 1835 * can be used by applications. 1836 */ 1837 if ((f2fs_compressed_file(inode) || f2fs_is_pinned_file(inode)) && 1838 (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE | 1839 FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE))) { 1840 ret = -EOPNOTSUPP; 1841 goto out; 1842 } 1843 1844 ret = file_modified(file); 1845 if (ret) 1846 goto out; 1847 1848 if (mode & FALLOC_FL_PUNCH_HOLE) { 1849 if (offset >= inode->i_size) 1850 goto out; 1851 1852 ret = f2fs_punch_hole(inode, offset, len); 1853 } else if (mode & FALLOC_FL_COLLAPSE_RANGE) { 1854 ret = f2fs_collapse_range(inode, offset, len); 1855 } else if (mode & FALLOC_FL_ZERO_RANGE) { 1856 ret = f2fs_zero_range(inode, offset, len, mode); 1857 } else if (mode & FALLOC_FL_INSERT_RANGE) { 1858 ret = f2fs_insert_range(inode, offset, len); 1859 } else { 1860 ret = f2fs_expand_inode_data(inode, offset, len, mode); 1861 } 1862 1863 if (!ret) { 1864 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 1865 f2fs_mark_inode_dirty_sync(inode, false); 1866 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 1867 } 1868 1869 out: 1870 inode_unlock(inode); 1871 1872 trace_f2fs_fallocate(inode, mode, offset, len, ret); 1873 return ret; 1874 } 1875 1876 static int f2fs_release_file(struct inode *inode, struct file *filp) 1877 { 1878 /* 1879 * f2fs_release_file is called at every close calls. So we should 1880 * not drop any inmemory pages by close called by other process. 1881 */ 1882 if (!(filp->f_mode & FMODE_WRITE) || 1883 atomic_read(&inode->i_writecount) != 1) 1884 return 0; 1885 1886 inode_lock(inode); 1887 f2fs_abort_atomic_write(inode, true); 1888 inode_unlock(inode); 1889 1890 return 0; 1891 } 1892 1893 static int f2fs_file_flush(struct file *file, fl_owner_t id) 1894 { 1895 struct inode *inode = file_inode(file); 1896 1897 /* 1898 * If the process doing a transaction is crashed, we should do 1899 * roll-back. Otherwise, other reader/write can see corrupted database 1900 * until all the writers close its file. Since this should be done 1901 * before dropping file lock, it needs to do in ->flush. 1902 */ 1903 if (F2FS_I(inode)->atomic_write_task == current && 1904 (current->flags & PF_EXITING)) { 1905 inode_lock(inode); 1906 f2fs_abort_atomic_write(inode, true); 1907 inode_unlock(inode); 1908 } 1909 1910 return 0; 1911 } 1912 1913 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask) 1914 { 1915 struct f2fs_inode_info *fi = F2FS_I(inode); 1916 u32 masked_flags = fi->i_flags & mask; 1917 1918 /* mask can be shrunk by flags_valid selector */ 1919 iflags &= mask; 1920 1921 /* Is it quota file? Do not allow user to mess with it */ 1922 if (IS_NOQUOTA(inode)) 1923 return -EPERM; 1924 1925 if ((iflags ^ masked_flags) & F2FS_CASEFOLD_FL) { 1926 if (!f2fs_sb_has_casefold(F2FS_I_SB(inode))) 1927 return -EOPNOTSUPP; 1928 if (!f2fs_empty_dir(inode)) 1929 return -ENOTEMPTY; 1930 } 1931 1932 if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) { 1933 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 1934 return -EOPNOTSUPP; 1935 if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL)) 1936 return -EINVAL; 1937 } 1938 1939 if ((iflags ^ masked_flags) & F2FS_COMPR_FL) { 1940 if (masked_flags & F2FS_COMPR_FL) { 1941 if (!f2fs_disable_compressed_file(inode)) 1942 return -EINVAL; 1943 } else { 1944 /* try to convert inline_data to support compression */ 1945 int err = f2fs_convert_inline_inode(inode); 1946 if (err) 1947 return err; 1948 1949 f2fs_down_write(&F2FS_I(inode)->i_sem); 1950 if (!f2fs_may_compress(inode) || 1951 (S_ISREG(inode->i_mode) && 1952 F2FS_HAS_BLOCKS(inode))) { 1953 f2fs_up_write(&F2FS_I(inode)->i_sem); 1954 return -EINVAL; 1955 } 1956 err = set_compress_context(inode); 1957 f2fs_up_write(&F2FS_I(inode)->i_sem); 1958 1959 if (err) 1960 return err; 1961 } 1962 } 1963 1964 fi->i_flags = iflags | (fi->i_flags & ~mask); 1965 f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) && 1966 (fi->i_flags & F2FS_NOCOMP_FL)); 1967 1968 if (fi->i_flags & F2FS_PROJINHERIT_FL) 1969 set_inode_flag(inode, FI_PROJ_INHERIT); 1970 else 1971 clear_inode_flag(inode, FI_PROJ_INHERIT); 1972 1973 inode_set_ctime_current(inode); 1974 f2fs_set_inode_flags(inode); 1975 f2fs_mark_inode_dirty_sync(inode, true); 1976 return 0; 1977 } 1978 1979 /* FS_IOC_[GS]ETFLAGS and FS_IOC_FS[GS]ETXATTR support */ 1980 1981 /* 1982 * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry 1983 * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to 1984 * F2FS_GETTABLE_FS_FL. To also make it settable via FS_IOC_SETFLAGS, also add 1985 * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL. 1986 * 1987 * Translating flags to fsx_flags value used by FS_IOC_FSGETXATTR and 1988 * FS_IOC_FSSETXATTR is done by the VFS. 1989 */ 1990 1991 static const struct { 1992 u32 iflag; 1993 u32 fsflag; 1994 } f2fs_fsflags_map[] = { 1995 { F2FS_COMPR_FL, FS_COMPR_FL }, 1996 { F2FS_SYNC_FL, FS_SYNC_FL }, 1997 { F2FS_IMMUTABLE_FL, FS_IMMUTABLE_FL }, 1998 { F2FS_APPEND_FL, FS_APPEND_FL }, 1999 { F2FS_NODUMP_FL, FS_NODUMP_FL }, 2000 { F2FS_NOATIME_FL, FS_NOATIME_FL }, 2001 { F2FS_NOCOMP_FL, FS_NOCOMP_FL }, 2002 { F2FS_INDEX_FL, FS_INDEX_FL }, 2003 { F2FS_DIRSYNC_FL, FS_DIRSYNC_FL }, 2004 { F2FS_PROJINHERIT_FL, FS_PROJINHERIT_FL }, 2005 { F2FS_CASEFOLD_FL, FS_CASEFOLD_FL }, 2006 }; 2007 2008 #define F2FS_GETTABLE_FS_FL ( \ 2009 FS_COMPR_FL | \ 2010 FS_SYNC_FL | \ 2011 FS_IMMUTABLE_FL | \ 2012 FS_APPEND_FL | \ 2013 FS_NODUMP_FL | \ 2014 FS_NOATIME_FL | \ 2015 FS_NOCOMP_FL | \ 2016 FS_INDEX_FL | \ 2017 FS_DIRSYNC_FL | \ 2018 FS_PROJINHERIT_FL | \ 2019 FS_ENCRYPT_FL | \ 2020 FS_INLINE_DATA_FL | \ 2021 FS_NOCOW_FL | \ 2022 FS_VERITY_FL | \ 2023 FS_CASEFOLD_FL) 2024 2025 #define F2FS_SETTABLE_FS_FL ( \ 2026 FS_COMPR_FL | \ 2027 FS_SYNC_FL | \ 2028 FS_IMMUTABLE_FL | \ 2029 FS_APPEND_FL | \ 2030 FS_NODUMP_FL | \ 2031 FS_NOATIME_FL | \ 2032 FS_NOCOMP_FL | \ 2033 FS_DIRSYNC_FL | \ 2034 FS_PROJINHERIT_FL | \ 2035 FS_CASEFOLD_FL) 2036 2037 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */ 2038 static inline u32 f2fs_iflags_to_fsflags(u32 iflags) 2039 { 2040 u32 fsflags = 0; 2041 int i; 2042 2043 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2044 if (iflags & f2fs_fsflags_map[i].iflag) 2045 fsflags |= f2fs_fsflags_map[i].fsflag; 2046 2047 return fsflags; 2048 } 2049 2050 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */ 2051 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags) 2052 { 2053 u32 iflags = 0; 2054 int i; 2055 2056 for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++) 2057 if (fsflags & f2fs_fsflags_map[i].fsflag) 2058 iflags |= f2fs_fsflags_map[i].iflag; 2059 2060 return iflags; 2061 } 2062 2063 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg) 2064 { 2065 struct inode *inode = file_inode(filp); 2066 2067 return put_user(inode->i_generation, (int __user *)arg); 2068 } 2069 2070 static int f2fs_ioc_start_atomic_write(struct file *filp, bool truncate) 2071 { 2072 struct inode *inode = file_inode(filp); 2073 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2074 struct f2fs_inode_info *fi = F2FS_I(inode); 2075 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2076 struct inode *pinode; 2077 loff_t isize; 2078 int ret; 2079 2080 if (!inode_owner_or_capable(idmap, inode)) 2081 return -EACCES; 2082 2083 if (!S_ISREG(inode->i_mode)) 2084 return -EINVAL; 2085 2086 if (filp->f_flags & O_DIRECT) 2087 return -EINVAL; 2088 2089 ret = mnt_want_write_file(filp); 2090 if (ret) 2091 return ret; 2092 2093 inode_lock(inode); 2094 2095 if (!f2fs_disable_compressed_file(inode) || 2096 f2fs_is_pinned_file(inode)) { 2097 ret = -EINVAL; 2098 goto out; 2099 } 2100 2101 if (f2fs_is_atomic_file(inode)) 2102 goto out; 2103 2104 ret = f2fs_convert_inline_inode(inode); 2105 if (ret) 2106 goto out; 2107 2108 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 2109 2110 /* 2111 * Should wait end_io to count F2FS_WB_CP_DATA correctly by 2112 * f2fs_is_atomic_file. 2113 */ 2114 if (get_dirty_pages(inode)) 2115 f2fs_warn(sbi, "Unexpected flush for atomic writes: ino=%lu, npages=%u", 2116 inode->i_ino, get_dirty_pages(inode)); 2117 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 2118 if (ret) { 2119 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2120 goto out; 2121 } 2122 2123 /* Check if the inode already has a COW inode */ 2124 if (fi->cow_inode == NULL) { 2125 /* Create a COW inode for atomic write */ 2126 pinode = f2fs_iget(inode->i_sb, fi->i_pino); 2127 if (IS_ERR(pinode)) { 2128 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2129 ret = PTR_ERR(pinode); 2130 goto out; 2131 } 2132 2133 ret = f2fs_get_tmpfile(idmap, pinode, &fi->cow_inode); 2134 iput(pinode); 2135 if (ret) { 2136 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2137 goto out; 2138 } 2139 2140 set_inode_flag(fi->cow_inode, FI_COW_FILE); 2141 clear_inode_flag(fi->cow_inode, FI_INLINE_DATA); 2142 } else { 2143 /* Reuse the already created COW inode */ 2144 ret = f2fs_do_truncate_blocks(fi->cow_inode, 0, true); 2145 if (ret) { 2146 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2147 goto out; 2148 } 2149 } 2150 2151 f2fs_write_inode(inode, NULL); 2152 2153 stat_inc_atomic_inode(inode); 2154 2155 set_inode_flag(inode, FI_ATOMIC_FILE); 2156 2157 isize = i_size_read(inode); 2158 fi->original_i_size = isize; 2159 if (truncate) { 2160 set_inode_flag(inode, FI_ATOMIC_REPLACE); 2161 truncate_inode_pages_final(inode->i_mapping); 2162 f2fs_i_size_write(inode, 0); 2163 isize = 0; 2164 } 2165 f2fs_i_size_write(fi->cow_inode, isize); 2166 2167 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 2168 2169 f2fs_update_time(sbi, REQ_TIME); 2170 fi->atomic_write_task = current; 2171 stat_update_max_atomic_write(inode); 2172 fi->atomic_write_cnt = 0; 2173 out: 2174 inode_unlock(inode); 2175 mnt_drop_write_file(filp); 2176 return ret; 2177 } 2178 2179 static int f2fs_ioc_commit_atomic_write(struct file *filp) 2180 { 2181 struct inode *inode = file_inode(filp); 2182 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2183 int ret; 2184 2185 if (!inode_owner_or_capable(idmap, inode)) 2186 return -EACCES; 2187 2188 ret = mnt_want_write_file(filp); 2189 if (ret) 2190 return ret; 2191 2192 f2fs_balance_fs(F2FS_I_SB(inode), true); 2193 2194 inode_lock(inode); 2195 2196 if (f2fs_is_atomic_file(inode)) { 2197 ret = f2fs_commit_atomic_write(inode); 2198 if (!ret) 2199 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true); 2200 2201 f2fs_abort_atomic_write(inode, ret); 2202 } else { 2203 ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false); 2204 } 2205 2206 inode_unlock(inode); 2207 mnt_drop_write_file(filp); 2208 return ret; 2209 } 2210 2211 static int f2fs_ioc_abort_atomic_write(struct file *filp) 2212 { 2213 struct inode *inode = file_inode(filp); 2214 struct mnt_idmap *idmap = file_mnt_idmap(filp); 2215 int ret; 2216 2217 if (!inode_owner_or_capable(idmap, inode)) 2218 return -EACCES; 2219 2220 ret = mnt_want_write_file(filp); 2221 if (ret) 2222 return ret; 2223 2224 inode_lock(inode); 2225 2226 f2fs_abort_atomic_write(inode, true); 2227 2228 inode_unlock(inode); 2229 2230 mnt_drop_write_file(filp); 2231 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2232 return ret; 2233 } 2234 2235 int f2fs_do_shutdown(struct f2fs_sb_info *sbi, unsigned int flag, 2236 bool readonly) 2237 { 2238 struct super_block *sb = sbi->sb; 2239 int ret = 0; 2240 2241 switch (flag) { 2242 case F2FS_GOING_DOWN_FULLSYNC: 2243 ret = bdev_freeze(sb->s_bdev); 2244 if (ret) 2245 goto out; 2246 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2247 bdev_thaw(sb->s_bdev); 2248 break; 2249 case F2FS_GOING_DOWN_METASYNC: 2250 /* do checkpoint only */ 2251 ret = f2fs_sync_fs(sb, 1); 2252 if (ret) { 2253 if (ret == -EIO) 2254 ret = 0; 2255 goto out; 2256 } 2257 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2258 break; 2259 case F2FS_GOING_DOWN_NOSYNC: 2260 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2261 break; 2262 case F2FS_GOING_DOWN_METAFLUSH: 2263 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO); 2264 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_SHUTDOWN); 2265 break; 2266 case F2FS_GOING_DOWN_NEED_FSCK: 2267 set_sbi_flag(sbi, SBI_NEED_FSCK); 2268 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 2269 set_sbi_flag(sbi, SBI_IS_DIRTY); 2270 /* do checkpoint only */ 2271 ret = f2fs_sync_fs(sb, 1); 2272 if (ret == -EIO) 2273 ret = 0; 2274 goto out; 2275 default: 2276 ret = -EINVAL; 2277 goto out; 2278 } 2279 2280 if (readonly) 2281 goto out; 2282 2283 f2fs_stop_gc_thread(sbi); 2284 f2fs_stop_discard_thread(sbi); 2285 2286 f2fs_drop_discard_cmd(sbi); 2287 clear_opt(sbi, DISCARD); 2288 2289 f2fs_update_time(sbi, REQ_TIME); 2290 out: 2291 2292 trace_f2fs_shutdown(sbi, flag, ret); 2293 2294 return ret; 2295 } 2296 2297 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg) 2298 { 2299 struct inode *inode = file_inode(filp); 2300 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2301 __u32 in; 2302 int ret; 2303 bool need_drop = false, readonly = false; 2304 2305 if (!capable(CAP_SYS_ADMIN)) 2306 return -EPERM; 2307 2308 if (get_user(in, (__u32 __user *)arg)) 2309 return -EFAULT; 2310 2311 if (in != F2FS_GOING_DOWN_FULLSYNC) { 2312 ret = mnt_want_write_file(filp); 2313 if (ret) { 2314 if (ret != -EROFS) 2315 return ret; 2316 2317 /* fallback to nosync shutdown for readonly fs */ 2318 in = F2FS_GOING_DOWN_NOSYNC; 2319 readonly = true; 2320 } else { 2321 need_drop = true; 2322 } 2323 } 2324 2325 ret = f2fs_do_shutdown(sbi, in, readonly); 2326 2327 if (need_drop) 2328 mnt_drop_write_file(filp); 2329 2330 return ret; 2331 } 2332 2333 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg) 2334 { 2335 struct inode *inode = file_inode(filp); 2336 struct super_block *sb = inode->i_sb; 2337 struct fstrim_range range; 2338 int ret; 2339 2340 if (!capable(CAP_SYS_ADMIN)) 2341 return -EPERM; 2342 2343 if (!f2fs_hw_support_discard(F2FS_SB(sb))) 2344 return -EOPNOTSUPP; 2345 2346 if (copy_from_user(&range, (struct fstrim_range __user *)arg, 2347 sizeof(range))) 2348 return -EFAULT; 2349 2350 ret = mnt_want_write_file(filp); 2351 if (ret) 2352 return ret; 2353 2354 range.minlen = max((unsigned int)range.minlen, 2355 bdev_discard_granularity(sb->s_bdev)); 2356 ret = f2fs_trim_fs(F2FS_SB(sb), &range); 2357 mnt_drop_write_file(filp); 2358 if (ret < 0) 2359 return ret; 2360 2361 if (copy_to_user((struct fstrim_range __user *)arg, &range, 2362 sizeof(range))) 2363 return -EFAULT; 2364 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2365 return 0; 2366 } 2367 2368 static bool uuid_is_nonzero(__u8 u[16]) 2369 { 2370 int i; 2371 2372 for (i = 0; i < 16; i++) 2373 if (u[i]) 2374 return true; 2375 return false; 2376 } 2377 2378 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg) 2379 { 2380 struct inode *inode = file_inode(filp); 2381 int ret; 2382 2383 if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode))) 2384 return -EOPNOTSUPP; 2385 2386 ret = fscrypt_ioctl_set_policy(filp, (const void __user *)arg); 2387 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 2388 return ret; 2389 } 2390 2391 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg) 2392 { 2393 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2394 return -EOPNOTSUPP; 2395 return fscrypt_ioctl_get_policy(filp, (void __user *)arg); 2396 } 2397 2398 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg) 2399 { 2400 struct inode *inode = file_inode(filp); 2401 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2402 u8 encrypt_pw_salt[16]; 2403 int err; 2404 2405 if (!f2fs_sb_has_encrypt(sbi)) 2406 return -EOPNOTSUPP; 2407 2408 err = mnt_want_write_file(filp); 2409 if (err) 2410 return err; 2411 2412 f2fs_down_write(&sbi->sb_lock); 2413 2414 if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt)) 2415 goto got_it; 2416 2417 /* update superblock with uuid */ 2418 generate_random_uuid(sbi->raw_super->encrypt_pw_salt); 2419 2420 err = f2fs_commit_super(sbi, false); 2421 if (err) { 2422 /* undo new data */ 2423 memset(sbi->raw_super->encrypt_pw_salt, 0, 16); 2424 goto out_err; 2425 } 2426 got_it: 2427 memcpy(encrypt_pw_salt, sbi->raw_super->encrypt_pw_salt, 16); 2428 out_err: 2429 f2fs_up_write(&sbi->sb_lock); 2430 mnt_drop_write_file(filp); 2431 2432 if (!err && copy_to_user((__u8 __user *)arg, encrypt_pw_salt, 16)) 2433 err = -EFAULT; 2434 2435 return err; 2436 } 2437 2438 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp, 2439 unsigned long arg) 2440 { 2441 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2442 return -EOPNOTSUPP; 2443 2444 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg); 2445 } 2446 2447 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg) 2448 { 2449 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2450 return -EOPNOTSUPP; 2451 2452 return fscrypt_ioctl_add_key(filp, (void __user *)arg); 2453 } 2454 2455 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg) 2456 { 2457 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2458 return -EOPNOTSUPP; 2459 2460 return fscrypt_ioctl_remove_key(filp, (void __user *)arg); 2461 } 2462 2463 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp, 2464 unsigned long arg) 2465 { 2466 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2467 return -EOPNOTSUPP; 2468 2469 return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg); 2470 } 2471 2472 static int f2fs_ioc_get_encryption_key_status(struct file *filp, 2473 unsigned long arg) 2474 { 2475 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2476 return -EOPNOTSUPP; 2477 2478 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg); 2479 } 2480 2481 static int f2fs_ioc_get_encryption_nonce(struct file *filp, unsigned long arg) 2482 { 2483 if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp)))) 2484 return -EOPNOTSUPP; 2485 2486 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg); 2487 } 2488 2489 static int f2fs_ioc_gc(struct file *filp, unsigned long arg) 2490 { 2491 struct inode *inode = file_inode(filp); 2492 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2493 struct f2fs_gc_control gc_control = { .victim_segno = NULL_SEGNO, 2494 .no_bg_gc = false, 2495 .should_migrate_blocks = false, 2496 .nr_free_secs = 0 }; 2497 __u32 sync; 2498 int ret; 2499 2500 if (!capable(CAP_SYS_ADMIN)) 2501 return -EPERM; 2502 2503 if (get_user(sync, (__u32 __user *)arg)) 2504 return -EFAULT; 2505 2506 if (f2fs_readonly(sbi->sb)) 2507 return -EROFS; 2508 2509 ret = mnt_want_write_file(filp); 2510 if (ret) 2511 return ret; 2512 2513 if (!sync) { 2514 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2515 ret = -EBUSY; 2516 goto out; 2517 } 2518 } else { 2519 f2fs_down_write(&sbi->gc_lock); 2520 } 2521 2522 gc_control.init_gc_type = sync ? FG_GC : BG_GC; 2523 gc_control.err_gc_skipped = sync; 2524 stat_inc_gc_call_count(sbi, FOREGROUND); 2525 ret = f2fs_gc(sbi, &gc_control); 2526 out: 2527 mnt_drop_write_file(filp); 2528 return ret; 2529 } 2530 2531 static int __f2fs_ioc_gc_range(struct file *filp, struct f2fs_gc_range *range) 2532 { 2533 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 2534 struct f2fs_gc_control gc_control = { 2535 .init_gc_type = range->sync ? FG_GC : BG_GC, 2536 .no_bg_gc = false, 2537 .should_migrate_blocks = false, 2538 .err_gc_skipped = range->sync, 2539 .nr_free_secs = 0 }; 2540 u64 end; 2541 int ret; 2542 2543 if (!capable(CAP_SYS_ADMIN)) 2544 return -EPERM; 2545 if (f2fs_readonly(sbi->sb)) 2546 return -EROFS; 2547 2548 end = range->start + range->len; 2549 if (end < range->start || range->start < MAIN_BLKADDR(sbi) || 2550 end >= MAX_BLKADDR(sbi)) 2551 return -EINVAL; 2552 2553 ret = mnt_want_write_file(filp); 2554 if (ret) 2555 return ret; 2556 2557 do_more: 2558 if (!range->sync) { 2559 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 2560 ret = -EBUSY; 2561 goto out; 2562 } 2563 } else { 2564 f2fs_down_write(&sbi->gc_lock); 2565 } 2566 2567 gc_control.victim_segno = GET_SEGNO(sbi, range->start); 2568 stat_inc_gc_call_count(sbi, FOREGROUND); 2569 ret = f2fs_gc(sbi, &gc_control); 2570 if (ret) { 2571 if (ret == -EBUSY) 2572 ret = -EAGAIN; 2573 goto out; 2574 } 2575 range->start += CAP_BLKS_PER_SEC(sbi); 2576 if (range->start <= end) 2577 goto do_more; 2578 out: 2579 mnt_drop_write_file(filp); 2580 return ret; 2581 } 2582 2583 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg) 2584 { 2585 struct f2fs_gc_range range; 2586 2587 if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg, 2588 sizeof(range))) 2589 return -EFAULT; 2590 return __f2fs_ioc_gc_range(filp, &range); 2591 } 2592 2593 static int f2fs_ioc_write_checkpoint(struct file *filp) 2594 { 2595 struct inode *inode = file_inode(filp); 2596 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2597 int ret; 2598 2599 if (!capable(CAP_SYS_ADMIN)) 2600 return -EPERM; 2601 2602 if (f2fs_readonly(sbi->sb)) 2603 return -EROFS; 2604 2605 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 2606 f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled."); 2607 return -EINVAL; 2608 } 2609 2610 ret = mnt_want_write_file(filp); 2611 if (ret) 2612 return ret; 2613 2614 ret = f2fs_sync_fs(sbi->sb, 1); 2615 2616 mnt_drop_write_file(filp); 2617 return ret; 2618 } 2619 2620 static int f2fs_defragment_range(struct f2fs_sb_info *sbi, 2621 struct file *filp, 2622 struct f2fs_defragment *range) 2623 { 2624 struct inode *inode = file_inode(filp); 2625 struct f2fs_map_blocks map = { .m_next_extent = NULL, 2626 .m_seg_type = NO_CHECK_TYPE, 2627 .m_may_create = false }; 2628 struct extent_info ei = {}; 2629 pgoff_t pg_start, pg_end, next_pgofs; 2630 unsigned int total = 0, sec_num; 2631 block_t blk_end = 0; 2632 bool fragmented = false; 2633 int err; 2634 2635 f2fs_balance_fs(sbi, true); 2636 2637 inode_lock(inode); 2638 pg_start = range->start >> PAGE_SHIFT; 2639 pg_end = min_t(pgoff_t, 2640 (range->start + range->len) >> PAGE_SHIFT, 2641 DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE)); 2642 2643 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 2644 err = -EINVAL; 2645 goto unlock_out; 2646 } 2647 2648 /* if in-place-update policy is enabled, don't waste time here */ 2649 set_inode_flag(inode, FI_OPU_WRITE); 2650 if (f2fs_should_update_inplace(inode, NULL)) { 2651 err = -EINVAL; 2652 goto out; 2653 } 2654 2655 /* writeback all dirty pages in the range */ 2656 err = filemap_write_and_wait_range(inode->i_mapping, 2657 pg_start << PAGE_SHIFT, 2658 (pg_end << PAGE_SHIFT) - 1); 2659 if (err) 2660 goto out; 2661 2662 /* 2663 * lookup mapping info in extent cache, skip defragmenting if physical 2664 * block addresses are continuous. 2665 */ 2666 if (f2fs_lookup_read_extent_cache(inode, pg_start, &ei)) { 2667 if (ei.fofs + ei.len >= pg_end) 2668 goto out; 2669 } 2670 2671 map.m_lblk = pg_start; 2672 map.m_next_pgofs = &next_pgofs; 2673 2674 /* 2675 * lookup mapping info in dnode page cache, skip defragmenting if all 2676 * physical block addresses are continuous even if there are hole(s) 2677 * in logical blocks. 2678 */ 2679 while (map.m_lblk < pg_end) { 2680 map.m_len = pg_end - map.m_lblk; 2681 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2682 if (err) 2683 goto out; 2684 2685 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2686 map.m_lblk = next_pgofs; 2687 continue; 2688 } 2689 2690 if (blk_end && blk_end != map.m_pblk) 2691 fragmented = true; 2692 2693 /* record total count of block that we're going to move */ 2694 total += map.m_len; 2695 2696 blk_end = map.m_pblk + map.m_len; 2697 2698 map.m_lblk += map.m_len; 2699 } 2700 2701 if (!fragmented) { 2702 total = 0; 2703 goto out; 2704 } 2705 2706 sec_num = DIV_ROUND_UP(total, CAP_BLKS_PER_SEC(sbi)); 2707 2708 /* 2709 * make sure there are enough free section for LFS allocation, this can 2710 * avoid defragment running in SSR mode when free section are allocated 2711 * intensively 2712 */ 2713 if (has_not_enough_free_secs(sbi, 0, sec_num)) { 2714 err = -EAGAIN; 2715 goto out; 2716 } 2717 2718 map.m_lblk = pg_start; 2719 map.m_len = pg_end - pg_start; 2720 total = 0; 2721 2722 while (map.m_lblk < pg_end) { 2723 pgoff_t idx; 2724 int cnt = 0; 2725 2726 do_map: 2727 map.m_len = pg_end - map.m_lblk; 2728 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT); 2729 if (err) 2730 goto clear_out; 2731 2732 if (!(map.m_flags & F2FS_MAP_FLAGS)) { 2733 map.m_lblk = next_pgofs; 2734 goto check; 2735 } 2736 2737 set_inode_flag(inode, FI_SKIP_WRITES); 2738 2739 idx = map.m_lblk; 2740 while (idx < map.m_lblk + map.m_len && 2741 cnt < BLKS_PER_SEG(sbi)) { 2742 struct page *page; 2743 2744 page = f2fs_get_lock_data_page(inode, idx, true); 2745 if (IS_ERR(page)) { 2746 err = PTR_ERR(page); 2747 goto clear_out; 2748 } 2749 2750 set_page_dirty(page); 2751 set_page_private_gcing(page); 2752 f2fs_put_page(page, 1); 2753 2754 idx++; 2755 cnt++; 2756 total++; 2757 } 2758 2759 map.m_lblk = idx; 2760 check: 2761 if (map.m_lblk < pg_end && cnt < BLKS_PER_SEG(sbi)) 2762 goto do_map; 2763 2764 clear_inode_flag(inode, FI_SKIP_WRITES); 2765 2766 err = filemap_fdatawrite(inode->i_mapping); 2767 if (err) 2768 goto out; 2769 } 2770 clear_out: 2771 clear_inode_flag(inode, FI_SKIP_WRITES); 2772 out: 2773 clear_inode_flag(inode, FI_OPU_WRITE); 2774 unlock_out: 2775 inode_unlock(inode); 2776 if (!err) 2777 range->len = (u64)total << PAGE_SHIFT; 2778 return err; 2779 } 2780 2781 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg) 2782 { 2783 struct inode *inode = file_inode(filp); 2784 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2785 struct f2fs_defragment range; 2786 int err; 2787 2788 if (!capable(CAP_SYS_ADMIN)) 2789 return -EPERM; 2790 2791 if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode)) 2792 return -EINVAL; 2793 2794 if (f2fs_readonly(sbi->sb)) 2795 return -EROFS; 2796 2797 if (copy_from_user(&range, (struct f2fs_defragment __user *)arg, 2798 sizeof(range))) 2799 return -EFAULT; 2800 2801 /* verify alignment of offset & size */ 2802 if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1)) 2803 return -EINVAL; 2804 2805 if (unlikely((range.start + range.len) >> PAGE_SHIFT > 2806 max_file_blocks(inode))) 2807 return -EINVAL; 2808 2809 err = mnt_want_write_file(filp); 2810 if (err) 2811 return err; 2812 2813 err = f2fs_defragment_range(sbi, filp, &range); 2814 mnt_drop_write_file(filp); 2815 2816 if (range.len) 2817 f2fs_update_time(sbi, REQ_TIME); 2818 if (err < 0) 2819 return err; 2820 2821 if (copy_to_user((struct f2fs_defragment __user *)arg, &range, 2822 sizeof(range))) 2823 return -EFAULT; 2824 2825 return 0; 2826 } 2827 2828 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in, 2829 struct file *file_out, loff_t pos_out, size_t len) 2830 { 2831 struct inode *src = file_inode(file_in); 2832 struct inode *dst = file_inode(file_out); 2833 struct f2fs_sb_info *sbi = F2FS_I_SB(src); 2834 size_t olen = len, dst_max_i_size = 0; 2835 size_t dst_osize; 2836 int ret; 2837 2838 if (file_in->f_path.mnt != file_out->f_path.mnt || 2839 src->i_sb != dst->i_sb) 2840 return -EXDEV; 2841 2842 if (unlikely(f2fs_readonly(src->i_sb))) 2843 return -EROFS; 2844 2845 if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode)) 2846 return -EINVAL; 2847 2848 if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst)) 2849 return -EOPNOTSUPP; 2850 2851 if (pos_out < 0 || pos_in < 0) 2852 return -EINVAL; 2853 2854 if (src == dst) { 2855 if (pos_in == pos_out) 2856 return 0; 2857 if (pos_out > pos_in && pos_out < pos_in + len) 2858 return -EINVAL; 2859 } 2860 2861 inode_lock(src); 2862 if (src != dst) { 2863 ret = -EBUSY; 2864 if (!inode_trylock(dst)) 2865 goto out; 2866 } 2867 2868 if (f2fs_compressed_file(src) || f2fs_compressed_file(dst) || 2869 f2fs_is_pinned_file(src) || f2fs_is_pinned_file(dst)) { 2870 ret = -EOPNOTSUPP; 2871 goto out_unlock; 2872 } 2873 2874 ret = -EINVAL; 2875 if (pos_in + len > src->i_size || pos_in + len < pos_in) 2876 goto out_unlock; 2877 if (len == 0) 2878 olen = len = src->i_size - pos_in; 2879 if (pos_in + len == src->i_size) 2880 len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in; 2881 if (len == 0) { 2882 ret = 0; 2883 goto out_unlock; 2884 } 2885 2886 dst_osize = dst->i_size; 2887 if (pos_out + olen > dst->i_size) 2888 dst_max_i_size = pos_out + olen; 2889 2890 /* verify the end result is block aligned */ 2891 if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) || 2892 !IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) || 2893 !IS_ALIGNED(pos_out, F2FS_BLKSIZE)) 2894 goto out_unlock; 2895 2896 ret = f2fs_convert_inline_inode(src); 2897 if (ret) 2898 goto out_unlock; 2899 2900 ret = f2fs_convert_inline_inode(dst); 2901 if (ret) 2902 goto out_unlock; 2903 2904 /* write out all dirty pages from offset */ 2905 ret = filemap_write_and_wait_range(src->i_mapping, 2906 pos_in, pos_in + len); 2907 if (ret) 2908 goto out_unlock; 2909 2910 ret = filemap_write_and_wait_range(dst->i_mapping, 2911 pos_out, pos_out + len); 2912 if (ret) 2913 goto out_unlock; 2914 2915 f2fs_balance_fs(sbi, true); 2916 2917 f2fs_down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2918 if (src != dst) { 2919 ret = -EBUSY; 2920 if (!f2fs_down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE])) 2921 goto out_src; 2922 } 2923 2924 f2fs_lock_op(sbi); 2925 ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS, 2926 pos_out >> F2FS_BLKSIZE_BITS, 2927 len >> F2FS_BLKSIZE_BITS, false); 2928 2929 if (!ret) { 2930 if (dst_max_i_size) 2931 f2fs_i_size_write(dst, dst_max_i_size); 2932 else if (dst_osize != dst->i_size) 2933 f2fs_i_size_write(dst, dst_osize); 2934 } 2935 f2fs_unlock_op(sbi); 2936 2937 if (src != dst) 2938 f2fs_up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]); 2939 out_src: 2940 f2fs_up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]); 2941 if (ret) 2942 goto out_unlock; 2943 2944 inode_set_mtime_to_ts(src, inode_set_ctime_current(src)); 2945 f2fs_mark_inode_dirty_sync(src, false); 2946 if (src != dst) { 2947 inode_set_mtime_to_ts(dst, inode_set_ctime_current(dst)); 2948 f2fs_mark_inode_dirty_sync(dst, false); 2949 } 2950 f2fs_update_time(sbi, REQ_TIME); 2951 2952 out_unlock: 2953 if (src != dst) 2954 inode_unlock(dst); 2955 out: 2956 inode_unlock(src); 2957 return ret; 2958 } 2959 2960 static int __f2fs_ioc_move_range(struct file *filp, 2961 struct f2fs_move_range *range) 2962 { 2963 struct fd dst; 2964 int err; 2965 2966 if (!(filp->f_mode & FMODE_READ) || 2967 !(filp->f_mode & FMODE_WRITE)) 2968 return -EBADF; 2969 2970 dst = fdget(range->dst_fd); 2971 if (!dst.file) 2972 return -EBADF; 2973 2974 if (!(dst.file->f_mode & FMODE_WRITE)) { 2975 err = -EBADF; 2976 goto err_out; 2977 } 2978 2979 err = mnt_want_write_file(filp); 2980 if (err) 2981 goto err_out; 2982 2983 err = f2fs_move_file_range(filp, range->pos_in, dst.file, 2984 range->pos_out, range->len); 2985 2986 mnt_drop_write_file(filp); 2987 err_out: 2988 fdput(dst); 2989 return err; 2990 } 2991 2992 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg) 2993 { 2994 struct f2fs_move_range range; 2995 2996 if (copy_from_user(&range, (struct f2fs_move_range __user *)arg, 2997 sizeof(range))) 2998 return -EFAULT; 2999 return __f2fs_ioc_move_range(filp, &range); 3000 } 3001 3002 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg) 3003 { 3004 struct inode *inode = file_inode(filp); 3005 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3006 struct sit_info *sm = SIT_I(sbi); 3007 unsigned int start_segno = 0, end_segno = 0; 3008 unsigned int dev_start_segno = 0, dev_end_segno = 0; 3009 struct f2fs_flush_device range; 3010 struct f2fs_gc_control gc_control = { 3011 .init_gc_type = FG_GC, 3012 .should_migrate_blocks = true, 3013 .err_gc_skipped = true, 3014 .nr_free_secs = 0 }; 3015 int ret; 3016 3017 if (!capable(CAP_SYS_ADMIN)) 3018 return -EPERM; 3019 3020 if (f2fs_readonly(sbi->sb)) 3021 return -EROFS; 3022 3023 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 3024 return -EINVAL; 3025 3026 if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg, 3027 sizeof(range))) 3028 return -EFAULT; 3029 3030 if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num || 3031 __is_large_section(sbi)) { 3032 f2fs_warn(sbi, "Can't flush %u in %d for SEGS_PER_SEC %u != 1", 3033 range.dev_num, sbi->s_ndevs, SEGS_PER_SEC(sbi)); 3034 return -EINVAL; 3035 } 3036 3037 ret = mnt_want_write_file(filp); 3038 if (ret) 3039 return ret; 3040 3041 if (range.dev_num != 0) 3042 dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk); 3043 dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk); 3044 3045 start_segno = sm->last_victim[FLUSH_DEVICE]; 3046 if (start_segno < dev_start_segno || start_segno >= dev_end_segno) 3047 start_segno = dev_start_segno; 3048 end_segno = min(start_segno + range.segments, dev_end_segno); 3049 3050 while (start_segno < end_segno) { 3051 if (!f2fs_down_write_trylock(&sbi->gc_lock)) { 3052 ret = -EBUSY; 3053 goto out; 3054 } 3055 sm->last_victim[GC_CB] = end_segno + 1; 3056 sm->last_victim[GC_GREEDY] = end_segno + 1; 3057 sm->last_victim[ALLOC_NEXT] = end_segno + 1; 3058 3059 gc_control.victim_segno = start_segno; 3060 stat_inc_gc_call_count(sbi, FOREGROUND); 3061 ret = f2fs_gc(sbi, &gc_control); 3062 if (ret == -EAGAIN) 3063 ret = 0; 3064 else if (ret < 0) 3065 break; 3066 start_segno++; 3067 } 3068 out: 3069 mnt_drop_write_file(filp); 3070 return ret; 3071 } 3072 3073 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg) 3074 { 3075 struct inode *inode = file_inode(filp); 3076 u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature); 3077 3078 /* Must validate to set it with SQLite behavior in Android. */ 3079 sb_feature |= F2FS_FEATURE_ATOMIC_WRITE; 3080 3081 return put_user(sb_feature, (u32 __user *)arg); 3082 } 3083 3084 #ifdef CONFIG_QUOTA 3085 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3086 { 3087 struct dquot *transfer_to[MAXQUOTAS] = {}; 3088 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3089 struct super_block *sb = sbi->sb; 3090 int err; 3091 3092 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid)); 3093 if (IS_ERR(transfer_to[PRJQUOTA])) 3094 return PTR_ERR(transfer_to[PRJQUOTA]); 3095 3096 err = __dquot_transfer(inode, transfer_to); 3097 if (err) 3098 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3099 dqput(transfer_to[PRJQUOTA]); 3100 return err; 3101 } 3102 3103 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3104 { 3105 struct f2fs_inode_info *fi = F2FS_I(inode); 3106 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3107 struct f2fs_inode *ri = NULL; 3108 kprojid_t kprojid; 3109 int err; 3110 3111 if (!f2fs_sb_has_project_quota(sbi)) { 3112 if (projid != F2FS_DEF_PROJID) 3113 return -EOPNOTSUPP; 3114 else 3115 return 0; 3116 } 3117 3118 if (!f2fs_has_extra_attr(inode)) 3119 return -EOPNOTSUPP; 3120 3121 kprojid = make_kprojid(&init_user_ns, (projid_t)projid); 3122 3123 if (projid_eq(kprojid, fi->i_projid)) 3124 return 0; 3125 3126 err = -EPERM; 3127 /* Is it quota file? Do not allow user to mess with it */ 3128 if (IS_NOQUOTA(inode)) 3129 return err; 3130 3131 if (!F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid)) 3132 return -EOVERFLOW; 3133 3134 err = f2fs_dquot_initialize(inode); 3135 if (err) 3136 return err; 3137 3138 f2fs_lock_op(sbi); 3139 err = f2fs_transfer_project_quota(inode, kprojid); 3140 if (err) 3141 goto out_unlock; 3142 3143 fi->i_projid = kprojid; 3144 inode_set_ctime_current(inode); 3145 f2fs_mark_inode_dirty_sync(inode, true); 3146 out_unlock: 3147 f2fs_unlock_op(sbi); 3148 return err; 3149 } 3150 #else 3151 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid) 3152 { 3153 return 0; 3154 } 3155 3156 static int f2fs_ioc_setproject(struct inode *inode, __u32 projid) 3157 { 3158 if (projid != F2FS_DEF_PROJID) 3159 return -EOPNOTSUPP; 3160 return 0; 3161 } 3162 #endif 3163 3164 int f2fs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 3165 { 3166 struct inode *inode = d_inode(dentry); 3167 struct f2fs_inode_info *fi = F2FS_I(inode); 3168 u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags); 3169 3170 if (IS_ENCRYPTED(inode)) 3171 fsflags |= FS_ENCRYPT_FL; 3172 if (IS_VERITY(inode)) 3173 fsflags |= FS_VERITY_FL; 3174 if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) 3175 fsflags |= FS_INLINE_DATA_FL; 3176 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3177 fsflags |= FS_NOCOW_FL; 3178 3179 fileattr_fill_flags(fa, fsflags & F2FS_GETTABLE_FS_FL); 3180 3181 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode))) 3182 fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid); 3183 3184 return 0; 3185 } 3186 3187 int f2fs_fileattr_set(struct mnt_idmap *idmap, 3188 struct dentry *dentry, struct fileattr *fa) 3189 { 3190 struct inode *inode = d_inode(dentry); 3191 u32 fsflags = fa->flags, mask = F2FS_SETTABLE_FS_FL; 3192 u32 iflags; 3193 int err; 3194 3195 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) 3196 return -EIO; 3197 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode))) 3198 return -ENOSPC; 3199 if (fsflags & ~F2FS_GETTABLE_FS_FL) 3200 return -EOPNOTSUPP; 3201 fsflags &= F2FS_SETTABLE_FS_FL; 3202 if (!fa->flags_valid) 3203 mask &= FS_COMMON_FL; 3204 3205 iflags = f2fs_fsflags_to_iflags(fsflags); 3206 if (f2fs_mask_flags(inode->i_mode, iflags) != iflags) 3207 return -EOPNOTSUPP; 3208 3209 err = f2fs_setflags_common(inode, iflags, f2fs_fsflags_to_iflags(mask)); 3210 if (!err) 3211 err = f2fs_ioc_setproject(inode, fa->fsx_projid); 3212 3213 return err; 3214 } 3215 3216 int f2fs_pin_file_control(struct inode *inode, bool inc) 3217 { 3218 struct f2fs_inode_info *fi = F2FS_I(inode); 3219 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3220 3221 if (fi->i_gc_failures >= sbi->gc_pin_file_threshold) { 3222 f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials", 3223 __func__, inode->i_ino, fi->i_gc_failures); 3224 clear_inode_flag(inode, FI_PIN_FILE); 3225 return -EAGAIN; 3226 } 3227 3228 /* Use i_gc_failures for normal file as a risk signal. */ 3229 if (inc) 3230 f2fs_i_gc_failures_write(inode, fi->i_gc_failures + 1); 3231 3232 return 0; 3233 } 3234 3235 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg) 3236 { 3237 struct inode *inode = file_inode(filp); 3238 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3239 __u32 pin; 3240 int ret = 0; 3241 3242 if (get_user(pin, (__u32 __user *)arg)) 3243 return -EFAULT; 3244 3245 if (!S_ISREG(inode->i_mode)) 3246 return -EINVAL; 3247 3248 if (f2fs_readonly(sbi->sb)) 3249 return -EROFS; 3250 3251 ret = mnt_want_write_file(filp); 3252 if (ret) 3253 return ret; 3254 3255 inode_lock(inode); 3256 3257 if (!pin) { 3258 clear_inode_flag(inode, FI_PIN_FILE); 3259 f2fs_i_gc_failures_write(inode, 0); 3260 goto done; 3261 } else if (f2fs_is_pinned_file(inode)) { 3262 goto done; 3263 } 3264 3265 if (F2FS_HAS_BLOCKS(inode)) { 3266 ret = -EFBIG; 3267 goto out; 3268 } 3269 3270 /* Let's allow file pinning on zoned device. */ 3271 if (!f2fs_sb_has_blkzoned(sbi) && 3272 f2fs_should_update_outplace(inode, NULL)) { 3273 ret = -EINVAL; 3274 goto out; 3275 } 3276 3277 if (f2fs_pin_file_control(inode, false)) { 3278 ret = -EAGAIN; 3279 goto out; 3280 } 3281 3282 ret = f2fs_convert_inline_inode(inode); 3283 if (ret) 3284 goto out; 3285 3286 if (!f2fs_disable_compressed_file(inode)) { 3287 ret = -EOPNOTSUPP; 3288 goto out; 3289 } 3290 3291 set_inode_flag(inode, FI_PIN_FILE); 3292 ret = F2FS_I(inode)->i_gc_failures; 3293 done: 3294 f2fs_update_time(sbi, REQ_TIME); 3295 out: 3296 inode_unlock(inode); 3297 mnt_drop_write_file(filp); 3298 return ret; 3299 } 3300 3301 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg) 3302 { 3303 struct inode *inode = file_inode(filp); 3304 __u32 pin = 0; 3305 3306 if (is_inode_flag_set(inode, FI_PIN_FILE)) 3307 pin = F2FS_I(inode)->i_gc_failures; 3308 return put_user(pin, (u32 __user *)arg); 3309 } 3310 3311 int f2fs_precache_extents(struct inode *inode) 3312 { 3313 struct f2fs_inode_info *fi = F2FS_I(inode); 3314 struct f2fs_map_blocks map; 3315 pgoff_t m_next_extent; 3316 loff_t end; 3317 int err; 3318 3319 if (is_inode_flag_set(inode, FI_NO_EXTENT)) 3320 return -EOPNOTSUPP; 3321 3322 map.m_lblk = 0; 3323 map.m_pblk = 0; 3324 map.m_next_pgofs = NULL; 3325 map.m_next_extent = &m_next_extent; 3326 map.m_seg_type = NO_CHECK_TYPE; 3327 map.m_may_create = false; 3328 end = F2FS_BLK_ALIGN(i_size_read(inode)); 3329 3330 while (map.m_lblk < end) { 3331 map.m_len = end - map.m_lblk; 3332 3333 f2fs_down_write(&fi->i_gc_rwsem[WRITE]); 3334 err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_PRECACHE); 3335 f2fs_up_write(&fi->i_gc_rwsem[WRITE]); 3336 if (err || !map.m_len) 3337 return err; 3338 3339 map.m_lblk = m_next_extent; 3340 } 3341 3342 return 0; 3343 } 3344 3345 static int f2fs_ioc_precache_extents(struct file *filp) 3346 { 3347 return f2fs_precache_extents(file_inode(filp)); 3348 } 3349 3350 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg) 3351 { 3352 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp)); 3353 __u64 block_count; 3354 3355 if (!capable(CAP_SYS_ADMIN)) 3356 return -EPERM; 3357 3358 if (f2fs_readonly(sbi->sb)) 3359 return -EROFS; 3360 3361 if (copy_from_user(&block_count, (void __user *)arg, 3362 sizeof(block_count))) 3363 return -EFAULT; 3364 3365 return f2fs_resize_fs(filp, block_count); 3366 } 3367 3368 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg) 3369 { 3370 struct inode *inode = file_inode(filp); 3371 3372 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); 3373 3374 if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) { 3375 f2fs_warn(F2FS_I_SB(inode), 3376 "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem", 3377 inode->i_ino); 3378 return -EOPNOTSUPP; 3379 } 3380 3381 return fsverity_ioctl_enable(filp, (const void __user *)arg); 3382 } 3383 3384 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg) 3385 { 3386 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3387 return -EOPNOTSUPP; 3388 3389 return fsverity_ioctl_measure(filp, (void __user *)arg); 3390 } 3391 3392 static int f2fs_ioc_read_verity_metadata(struct file *filp, unsigned long arg) 3393 { 3394 if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp)))) 3395 return -EOPNOTSUPP; 3396 3397 return fsverity_ioctl_read_metadata(filp, (const void __user *)arg); 3398 } 3399 3400 static int f2fs_ioc_getfslabel(struct file *filp, unsigned long arg) 3401 { 3402 struct inode *inode = file_inode(filp); 3403 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3404 char *vbuf; 3405 int count; 3406 int err = 0; 3407 3408 vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL); 3409 if (!vbuf) 3410 return -ENOMEM; 3411 3412 f2fs_down_read(&sbi->sb_lock); 3413 count = utf16s_to_utf8s(sbi->raw_super->volume_name, 3414 ARRAY_SIZE(sbi->raw_super->volume_name), 3415 UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME); 3416 f2fs_up_read(&sbi->sb_lock); 3417 3418 if (copy_to_user((char __user *)arg, vbuf, 3419 min(FSLABEL_MAX, count))) 3420 err = -EFAULT; 3421 3422 kfree(vbuf); 3423 return err; 3424 } 3425 3426 static int f2fs_ioc_setfslabel(struct file *filp, unsigned long arg) 3427 { 3428 struct inode *inode = file_inode(filp); 3429 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3430 char *vbuf; 3431 int err = 0; 3432 3433 if (!capable(CAP_SYS_ADMIN)) 3434 return -EPERM; 3435 3436 vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX); 3437 if (IS_ERR(vbuf)) 3438 return PTR_ERR(vbuf); 3439 3440 err = mnt_want_write_file(filp); 3441 if (err) 3442 goto out; 3443 3444 f2fs_down_write(&sbi->sb_lock); 3445 3446 memset(sbi->raw_super->volume_name, 0, 3447 sizeof(sbi->raw_super->volume_name)); 3448 utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN, 3449 sbi->raw_super->volume_name, 3450 ARRAY_SIZE(sbi->raw_super->volume_name)); 3451 3452 err = f2fs_commit_super(sbi, false); 3453 3454 f2fs_up_write(&sbi->sb_lock); 3455 3456 mnt_drop_write_file(filp); 3457 out: 3458 kfree(vbuf); 3459 return err; 3460 } 3461 3462 static int f2fs_get_compress_blocks(struct inode *inode, __u64 *blocks) 3463 { 3464 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 3465 return -EOPNOTSUPP; 3466 3467 if (!f2fs_compressed_file(inode)) 3468 return -EINVAL; 3469 3470 *blocks = atomic_read(&F2FS_I(inode)->i_compr_blocks); 3471 3472 return 0; 3473 } 3474 3475 static int f2fs_ioc_get_compress_blocks(struct file *filp, unsigned long arg) 3476 { 3477 struct inode *inode = file_inode(filp); 3478 __u64 blocks; 3479 int ret; 3480 3481 ret = f2fs_get_compress_blocks(inode, &blocks); 3482 if (ret < 0) 3483 return ret; 3484 3485 return put_user(blocks, (u64 __user *)arg); 3486 } 3487 3488 static int release_compress_blocks(struct dnode_of_data *dn, pgoff_t count) 3489 { 3490 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3491 unsigned int released_blocks = 0; 3492 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3493 block_t blkaddr; 3494 int i; 3495 3496 for (i = 0; i < count; i++) { 3497 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3498 dn->ofs_in_node + i); 3499 3500 if (!__is_valid_data_blkaddr(blkaddr)) 3501 continue; 3502 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3503 DATA_GENERIC_ENHANCE))) 3504 return -EFSCORRUPTED; 3505 } 3506 3507 while (count) { 3508 int compr_blocks = 0; 3509 3510 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3511 blkaddr = f2fs_data_blkaddr(dn); 3512 3513 if (i == 0) { 3514 if (blkaddr == COMPRESS_ADDR) 3515 continue; 3516 dn->ofs_in_node += cluster_size; 3517 goto next; 3518 } 3519 3520 if (__is_valid_data_blkaddr(blkaddr)) 3521 compr_blocks++; 3522 3523 if (blkaddr != NEW_ADDR) 3524 continue; 3525 3526 f2fs_set_data_blkaddr(dn, NULL_ADDR); 3527 } 3528 3529 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, false); 3530 dec_valid_block_count(sbi, dn->inode, 3531 cluster_size - compr_blocks); 3532 3533 released_blocks += cluster_size - compr_blocks; 3534 next: 3535 count -= cluster_size; 3536 } 3537 3538 return released_blocks; 3539 } 3540 3541 static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg) 3542 { 3543 struct inode *inode = file_inode(filp); 3544 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3545 pgoff_t page_idx = 0, last_idx; 3546 unsigned int released_blocks = 0; 3547 int ret; 3548 int writecount; 3549 3550 if (!f2fs_sb_has_compression(sbi)) 3551 return -EOPNOTSUPP; 3552 3553 if (f2fs_readonly(sbi->sb)) 3554 return -EROFS; 3555 3556 ret = mnt_want_write_file(filp); 3557 if (ret) 3558 return ret; 3559 3560 f2fs_balance_fs(sbi, true); 3561 3562 inode_lock(inode); 3563 3564 writecount = atomic_read(&inode->i_writecount); 3565 if ((filp->f_mode & FMODE_WRITE && writecount != 1) || 3566 (!(filp->f_mode & FMODE_WRITE) && writecount)) { 3567 ret = -EBUSY; 3568 goto out; 3569 } 3570 3571 if (!f2fs_compressed_file(inode) || 3572 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3573 ret = -EINVAL; 3574 goto out; 3575 } 3576 3577 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 3578 if (ret) 3579 goto out; 3580 3581 if (!atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3582 ret = -EPERM; 3583 goto out; 3584 } 3585 3586 set_inode_flag(inode, FI_COMPRESS_RELEASED); 3587 inode_set_ctime_current(inode); 3588 f2fs_mark_inode_dirty_sync(inode, true); 3589 3590 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3591 filemap_invalidate_lock(inode->i_mapping); 3592 3593 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3594 3595 while (page_idx < last_idx) { 3596 struct dnode_of_data dn; 3597 pgoff_t end_offset, count; 3598 3599 f2fs_lock_op(sbi); 3600 3601 set_new_dnode(&dn, inode, NULL, NULL, 0); 3602 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3603 if (ret) { 3604 f2fs_unlock_op(sbi); 3605 if (ret == -ENOENT) { 3606 page_idx = f2fs_get_next_page_offset(&dn, 3607 page_idx); 3608 ret = 0; 3609 continue; 3610 } 3611 break; 3612 } 3613 3614 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3615 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3616 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3617 3618 ret = release_compress_blocks(&dn, count); 3619 3620 f2fs_put_dnode(&dn); 3621 3622 f2fs_unlock_op(sbi); 3623 3624 if (ret < 0) 3625 break; 3626 3627 page_idx += count; 3628 released_blocks += ret; 3629 } 3630 3631 filemap_invalidate_unlock(inode->i_mapping); 3632 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3633 out: 3634 if (released_blocks) 3635 f2fs_update_time(sbi, REQ_TIME); 3636 inode_unlock(inode); 3637 3638 mnt_drop_write_file(filp); 3639 3640 if (ret >= 0) { 3641 ret = put_user(released_blocks, (u64 __user *)arg); 3642 } else if (released_blocks && 3643 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3644 set_sbi_flag(sbi, SBI_NEED_FSCK); 3645 f2fs_warn(sbi, "%s: partial blocks were released i_ino=%lx " 3646 "iblocks=%llu, released=%u, compr_blocks=%u, " 3647 "run fsck to fix.", 3648 __func__, inode->i_ino, inode->i_blocks, 3649 released_blocks, 3650 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3651 } 3652 3653 return ret; 3654 } 3655 3656 static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count, 3657 unsigned int *reserved_blocks) 3658 { 3659 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); 3660 int cluster_size = F2FS_I(dn->inode)->i_cluster_size; 3661 block_t blkaddr; 3662 int i; 3663 3664 for (i = 0; i < count; i++) { 3665 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3666 dn->ofs_in_node + i); 3667 3668 if (!__is_valid_data_blkaddr(blkaddr)) 3669 continue; 3670 if (unlikely(!f2fs_is_valid_blkaddr(sbi, blkaddr, 3671 DATA_GENERIC_ENHANCE))) 3672 return -EFSCORRUPTED; 3673 } 3674 3675 while (count) { 3676 int compr_blocks = 0; 3677 blkcnt_t reserved = 0; 3678 blkcnt_t to_reserved; 3679 int ret; 3680 3681 for (i = 0; i < cluster_size; i++) { 3682 blkaddr = data_blkaddr(dn->inode, dn->node_page, 3683 dn->ofs_in_node + i); 3684 3685 if (i == 0) { 3686 if (blkaddr != COMPRESS_ADDR) { 3687 dn->ofs_in_node += cluster_size; 3688 goto next; 3689 } 3690 continue; 3691 } 3692 3693 /* 3694 * compressed cluster was not released due to it 3695 * fails in release_compress_blocks(), so NEW_ADDR 3696 * is a possible case. 3697 */ 3698 if (blkaddr == NEW_ADDR) { 3699 reserved++; 3700 continue; 3701 } 3702 if (__is_valid_data_blkaddr(blkaddr)) { 3703 compr_blocks++; 3704 continue; 3705 } 3706 } 3707 3708 to_reserved = cluster_size - compr_blocks - reserved; 3709 3710 /* for the case all blocks in cluster were reserved */ 3711 if (to_reserved == 1) { 3712 dn->ofs_in_node += cluster_size; 3713 goto next; 3714 } 3715 3716 ret = inc_valid_block_count(sbi, dn->inode, 3717 &to_reserved, false); 3718 if (unlikely(ret)) 3719 return ret; 3720 3721 for (i = 0; i < cluster_size; i++, dn->ofs_in_node++) { 3722 if (f2fs_data_blkaddr(dn) == NULL_ADDR) 3723 f2fs_set_data_blkaddr(dn, NEW_ADDR); 3724 } 3725 3726 f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true); 3727 3728 *reserved_blocks += to_reserved; 3729 next: 3730 count -= cluster_size; 3731 } 3732 3733 return 0; 3734 } 3735 3736 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg) 3737 { 3738 struct inode *inode = file_inode(filp); 3739 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3740 pgoff_t page_idx = 0, last_idx; 3741 unsigned int reserved_blocks = 0; 3742 int ret; 3743 3744 if (!f2fs_sb_has_compression(sbi)) 3745 return -EOPNOTSUPP; 3746 3747 if (f2fs_readonly(sbi->sb)) 3748 return -EROFS; 3749 3750 ret = mnt_want_write_file(filp); 3751 if (ret) 3752 return ret; 3753 3754 f2fs_balance_fs(sbi, true); 3755 3756 inode_lock(inode); 3757 3758 if (!f2fs_compressed_file(inode) || 3759 !is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 3760 ret = -EINVAL; 3761 goto unlock_inode; 3762 } 3763 3764 if (atomic_read(&F2FS_I(inode)->i_compr_blocks)) 3765 goto unlock_inode; 3766 3767 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3768 filemap_invalidate_lock(inode->i_mapping); 3769 3770 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 3771 3772 while (page_idx < last_idx) { 3773 struct dnode_of_data dn; 3774 pgoff_t end_offset, count; 3775 3776 f2fs_lock_op(sbi); 3777 3778 set_new_dnode(&dn, inode, NULL, NULL, 0); 3779 ret = f2fs_get_dnode_of_data(&dn, page_idx, LOOKUP_NODE); 3780 if (ret) { 3781 f2fs_unlock_op(sbi); 3782 if (ret == -ENOENT) { 3783 page_idx = f2fs_get_next_page_offset(&dn, 3784 page_idx); 3785 ret = 0; 3786 continue; 3787 } 3788 break; 3789 } 3790 3791 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3792 count = min(end_offset - dn.ofs_in_node, last_idx - page_idx); 3793 count = round_up(count, F2FS_I(inode)->i_cluster_size); 3794 3795 ret = reserve_compress_blocks(&dn, count, &reserved_blocks); 3796 3797 f2fs_put_dnode(&dn); 3798 3799 f2fs_unlock_op(sbi); 3800 3801 if (ret < 0) 3802 break; 3803 3804 page_idx += count; 3805 } 3806 3807 filemap_invalidate_unlock(inode->i_mapping); 3808 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3809 3810 if (!ret) { 3811 clear_inode_flag(inode, FI_COMPRESS_RELEASED); 3812 inode_set_ctime_current(inode); 3813 f2fs_mark_inode_dirty_sync(inode, true); 3814 } 3815 unlock_inode: 3816 if (reserved_blocks) 3817 f2fs_update_time(sbi, REQ_TIME); 3818 inode_unlock(inode); 3819 mnt_drop_write_file(filp); 3820 3821 if (!ret) { 3822 ret = put_user(reserved_blocks, (u64 __user *)arg); 3823 } else if (reserved_blocks && 3824 atomic_read(&F2FS_I(inode)->i_compr_blocks)) { 3825 set_sbi_flag(sbi, SBI_NEED_FSCK); 3826 f2fs_warn(sbi, "%s: partial blocks were reserved i_ino=%lx " 3827 "iblocks=%llu, reserved=%u, compr_blocks=%u, " 3828 "run fsck to fix.", 3829 __func__, inode->i_ino, inode->i_blocks, 3830 reserved_blocks, 3831 atomic_read(&F2FS_I(inode)->i_compr_blocks)); 3832 } 3833 3834 return ret; 3835 } 3836 3837 static int f2fs_secure_erase(struct block_device *bdev, struct inode *inode, 3838 pgoff_t off, block_t block, block_t len, u32 flags) 3839 { 3840 sector_t sector = SECTOR_FROM_BLOCK(block); 3841 sector_t nr_sects = SECTOR_FROM_BLOCK(len); 3842 int ret = 0; 3843 3844 if (flags & F2FS_TRIM_FILE_DISCARD) { 3845 if (bdev_max_secure_erase_sectors(bdev)) 3846 ret = blkdev_issue_secure_erase(bdev, sector, nr_sects, 3847 GFP_NOFS); 3848 else 3849 ret = blkdev_issue_discard(bdev, sector, nr_sects, 3850 GFP_NOFS); 3851 } 3852 3853 if (!ret && (flags & F2FS_TRIM_FILE_ZEROOUT)) { 3854 if (IS_ENCRYPTED(inode)) 3855 ret = fscrypt_zeroout_range(inode, off, block, len); 3856 else 3857 ret = blkdev_issue_zeroout(bdev, sector, nr_sects, 3858 GFP_NOFS, 0); 3859 } 3860 3861 return ret; 3862 } 3863 3864 static int f2fs_sec_trim_file(struct file *filp, unsigned long arg) 3865 { 3866 struct inode *inode = file_inode(filp); 3867 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3868 struct address_space *mapping = inode->i_mapping; 3869 struct block_device *prev_bdev = NULL; 3870 struct f2fs_sectrim_range range; 3871 pgoff_t index, pg_end, prev_index = 0; 3872 block_t prev_block = 0, len = 0; 3873 loff_t end_addr; 3874 bool to_end = false; 3875 int ret = 0; 3876 3877 if (!(filp->f_mode & FMODE_WRITE)) 3878 return -EBADF; 3879 3880 if (copy_from_user(&range, (struct f2fs_sectrim_range __user *)arg, 3881 sizeof(range))) 3882 return -EFAULT; 3883 3884 if (range.flags == 0 || (range.flags & ~F2FS_TRIM_FILE_MASK) || 3885 !S_ISREG(inode->i_mode)) 3886 return -EINVAL; 3887 3888 if (((range.flags & F2FS_TRIM_FILE_DISCARD) && 3889 !f2fs_hw_support_discard(sbi)) || 3890 ((range.flags & F2FS_TRIM_FILE_ZEROOUT) && 3891 IS_ENCRYPTED(inode) && f2fs_is_multi_device(sbi))) 3892 return -EOPNOTSUPP; 3893 3894 file_start_write(filp); 3895 inode_lock(inode); 3896 3897 if (f2fs_is_atomic_file(inode) || f2fs_compressed_file(inode) || 3898 range.start >= inode->i_size) { 3899 ret = -EINVAL; 3900 goto err; 3901 } 3902 3903 if (range.len == 0) 3904 goto err; 3905 3906 if (inode->i_size - range.start > range.len) { 3907 end_addr = range.start + range.len; 3908 } else { 3909 end_addr = range.len == (u64)-1 ? 3910 sbi->sb->s_maxbytes : inode->i_size; 3911 to_end = true; 3912 } 3913 3914 if (!IS_ALIGNED(range.start, F2FS_BLKSIZE) || 3915 (!to_end && !IS_ALIGNED(end_addr, F2FS_BLKSIZE))) { 3916 ret = -EINVAL; 3917 goto err; 3918 } 3919 3920 index = F2FS_BYTES_TO_BLK(range.start); 3921 pg_end = DIV_ROUND_UP(end_addr, F2FS_BLKSIZE); 3922 3923 ret = f2fs_convert_inline_inode(inode); 3924 if (ret) 3925 goto err; 3926 3927 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 3928 filemap_invalidate_lock(mapping); 3929 3930 ret = filemap_write_and_wait_range(mapping, range.start, 3931 to_end ? LLONG_MAX : end_addr - 1); 3932 if (ret) 3933 goto out; 3934 3935 truncate_inode_pages_range(mapping, range.start, 3936 to_end ? -1 : end_addr - 1); 3937 3938 while (index < pg_end) { 3939 struct dnode_of_data dn; 3940 pgoff_t end_offset, count; 3941 int i; 3942 3943 set_new_dnode(&dn, inode, NULL, NULL, 0); 3944 ret = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE); 3945 if (ret) { 3946 if (ret == -ENOENT) { 3947 index = f2fs_get_next_page_offset(&dn, index); 3948 continue; 3949 } 3950 goto out; 3951 } 3952 3953 end_offset = ADDRS_PER_PAGE(dn.node_page, inode); 3954 count = min(end_offset - dn.ofs_in_node, pg_end - index); 3955 for (i = 0; i < count; i++, index++, dn.ofs_in_node++) { 3956 struct block_device *cur_bdev; 3957 block_t blkaddr = f2fs_data_blkaddr(&dn); 3958 3959 if (!__is_valid_data_blkaddr(blkaddr)) 3960 continue; 3961 3962 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, 3963 DATA_GENERIC_ENHANCE)) { 3964 ret = -EFSCORRUPTED; 3965 f2fs_put_dnode(&dn); 3966 goto out; 3967 } 3968 3969 cur_bdev = f2fs_target_device(sbi, blkaddr, NULL); 3970 if (f2fs_is_multi_device(sbi)) { 3971 int di = f2fs_target_device_index(sbi, blkaddr); 3972 3973 blkaddr -= FDEV(di).start_blk; 3974 } 3975 3976 if (len) { 3977 if (prev_bdev == cur_bdev && 3978 index == prev_index + len && 3979 blkaddr == prev_block + len) { 3980 len++; 3981 } else { 3982 ret = f2fs_secure_erase(prev_bdev, 3983 inode, prev_index, prev_block, 3984 len, range.flags); 3985 if (ret) { 3986 f2fs_put_dnode(&dn); 3987 goto out; 3988 } 3989 3990 len = 0; 3991 } 3992 } 3993 3994 if (!len) { 3995 prev_bdev = cur_bdev; 3996 prev_index = index; 3997 prev_block = blkaddr; 3998 len = 1; 3999 } 4000 } 4001 4002 f2fs_put_dnode(&dn); 4003 4004 if (fatal_signal_pending(current)) { 4005 ret = -EINTR; 4006 goto out; 4007 } 4008 cond_resched(); 4009 } 4010 4011 if (len) 4012 ret = f2fs_secure_erase(prev_bdev, inode, prev_index, 4013 prev_block, len, range.flags); 4014 f2fs_update_time(sbi, REQ_TIME); 4015 out: 4016 filemap_invalidate_unlock(mapping); 4017 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4018 err: 4019 inode_unlock(inode); 4020 file_end_write(filp); 4021 4022 return ret; 4023 } 4024 4025 static int f2fs_ioc_get_compress_option(struct file *filp, unsigned long arg) 4026 { 4027 struct inode *inode = file_inode(filp); 4028 struct f2fs_comp_option option; 4029 4030 if (!f2fs_sb_has_compression(F2FS_I_SB(inode))) 4031 return -EOPNOTSUPP; 4032 4033 inode_lock_shared(inode); 4034 4035 if (!f2fs_compressed_file(inode)) { 4036 inode_unlock_shared(inode); 4037 return -ENODATA; 4038 } 4039 4040 option.algorithm = F2FS_I(inode)->i_compress_algorithm; 4041 option.log_cluster_size = F2FS_I(inode)->i_log_cluster_size; 4042 4043 inode_unlock_shared(inode); 4044 4045 if (copy_to_user((struct f2fs_comp_option __user *)arg, &option, 4046 sizeof(option))) 4047 return -EFAULT; 4048 4049 return 0; 4050 } 4051 4052 static int f2fs_ioc_set_compress_option(struct file *filp, unsigned long arg) 4053 { 4054 struct inode *inode = file_inode(filp); 4055 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4056 struct f2fs_comp_option option; 4057 int ret = 0; 4058 4059 if (!f2fs_sb_has_compression(sbi)) 4060 return -EOPNOTSUPP; 4061 4062 if (!(filp->f_mode & FMODE_WRITE)) 4063 return -EBADF; 4064 4065 if (copy_from_user(&option, (struct f2fs_comp_option __user *)arg, 4066 sizeof(option))) 4067 return -EFAULT; 4068 4069 if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE || 4070 option.log_cluster_size > MAX_COMPRESS_LOG_SIZE || 4071 option.algorithm >= COMPRESS_MAX) 4072 return -EINVAL; 4073 4074 file_start_write(filp); 4075 inode_lock(inode); 4076 4077 f2fs_down_write(&F2FS_I(inode)->i_sem); 4078 if (!f2fs_compressed_file(inode)) { 4079 ret = -EINVAL; 4080 goto out; 4081 } 4082 4083 if (f2fs_is_mmap_file(inode) || get_dirty_pages(inode)) { 4084 ret = -EBUSY; 4085 goto out; 4086 } 4087 4088 if (F2FS_HAS_BLOCKS(inode)) { 4089 ret = -EFBIG; 4090 goto out; 4091 } 4092 4093 F2FS_I(inode)->i_compress_algorithm = option.algorithm; 4094 F2FS_I(inode)->i_log_cluster_size = option.log_cluster_size; 4095 F2FS_I(inode)->i_cluster_size = BIT(option.log_cluster_size); 4096 /* Set default level */ 4097 if (F2FS_I(inode)->i_compress_algorithm == COMPRESS_ZSTD) 4098 F2FS_I(inode)->i_compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 4099 else 4100 F2FS_I(inode)->i_compress_level = 0; 4101 /* Adjust mount option level */ 4102 if (option.algorithm == F2FS_OPTION(sbi).compress_algorithm && 4103 F2FS_OPTION(sbi).compress_level) 4104 F2FS_I(inode)->i_compress_level = F2FS_OPTION(sbi).compress_level; 4105 f2fs_mark_inode_dirty_sync(inode, true); 4106 4107 if (!f2fs_is_compress_backend_ready(inode)) 4108 f2fs_warn(sbi, "compression algorithm is successfully set, " 4109 "but current kernel doesn't support this algorithm."); 4110 out: 4111 f2fs_up_write(&F2FS_I(inode)->i_sem); 4112 inode_unlock(inode); 4113 file_end_write(filp); 4114 4115 return ret; 4116 } 4117 4118 static int redirty_blocks(struct inode *inode, pgoff_t page_idx, int len) 4119 { 4120 DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, page_idx); 4121 struct address_space *mapping = inode->i_mapping; 4122 struct page *page; 4123 pgoff_t redirty_idx = page_idx; 4124 int i, page_len = 0, ret = 0; 4125 4126 page_cache_ra_unbounded(&ractl, len, 0); 4127 4128 for (i = 0; i < len; i++, page_idx++) { 4129 page = read_cache_page(mapping, page_idx, NULL, NULL); 4130 if (IS_ERR(page)) { 4131 ret = PTR_ERR(page); 4132 break; 4133 } 4134 page_len++; 4135 } 4136 4137 for (i = 0; i < page_len; i++, redirty_idx++) { 4138 page = find_lock_page(mapping, redirty_idx); 4139 4140 /* It will never fail, when page has pinned above */ 4141 f2fs_bug_on(F2FS_I_SB(inode), !page); 4142 4143 set_page_dirty(page); 4144 set_page_private_gcing(page); 4145 f2fs_put_page(page, 1); 4146 f2fs_put_page(page, 0); 4147 } 4148 4149 return ret; 4150 } 4151 4152 static int f2fs_ioc_decompress_file(struct file *filp) 4153 { 4154 struct inode *inode = file_inode(filp); 4155 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4156 struct f2fs_inode_info *fi = F2FS_I(inode); 4157 pgoff_t page_idx = 0, last_idx; 4158 int cluster_size = fi->i_cluster_size; 4159 int count, ret; 4160 4161 if (!f2fs_sb_has_compression(sbi) || 4162 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4163 return -EOPNOTSUPP; 4164 4165 if (!(filp->f_mode & FMODE_WRITE)) 4166 return -EBADF; 4167 4168 f2fs_balance_fs(sbi, true); 4169 4170 file_start_write(filp); 4171 inode_lock(inode); 4172 4173 if (!f2fs_is_compress_backend_ready(inode)) { 4174 ret = -EOPNOTSUPP; 4175 goto out; 4176 } 4177 4178 if (!f2fs_compressed_file(inode) || 4179 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4180 ret = -EINVAL; 4181 goto out; 4182 } 4183 4184 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4185 if (ret) 4186 goto out; 4187 4188 if (!atomic_read(&fi->i_compr_blocks)) 4189 goto out; 4190 4191 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4192 4193 count = last_idx - page_idx; 4194 while (count && count >= cluster_size) { 4195 ret = redirty_blocks(inode, page_idx, cluster_size); 4196 if (ret < 0) 4197 break; 4198 4199 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4200 ret = filemap_fdatawrite(inode->i_mapping); 4201 if (ret < 0) 4202 break; 4203 } 4204 4205 count -= cluster_size; 4206 page_idx += cluster_size; 4207 4208 cond_resched(); 4209 if (fatal_signal_pending(current)) { 4210 ret = -EINTR; 4211 break; 4212 } 4213 } 4214 4215 if (!ret) 4216 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4217 LLONG_MAX); 4218 4219 if (ret) 4220 f2fs_warn(sbi, "%s: The file might be partially decompressed (errno=%d). Please delete the file.", 4221 __func__, ret); 4222 f2fs_update_time(sbi, REQ_TIME); 4223 out: 4224 inode_unlock(inode); 4225 file_end_write(filp); 4226 4227 return ret; 4228 } 4229 4230 static int f2fs_ioc_compress_file(struct file *filp) 4231 { 4232 struct inode *inode = file_inode(filp); 4233 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4234 pgoff_t page_idx = 0, last_idx; 4235 int cluster_size = F2FS_I(inode)->i_cluster_size; 4236 int count, ret; 4237 4238 if (!f2fs_sb_has_compression(sbi) || 4239 F2FS_OPTION(sbi).compress_mode != COMPR_MODE_USER) 4240 return -EOPNOTSUPP; 4241 4242 if (!(filp->f_mode & FMODE_WRITE)) 4243 return -EBADF; 4244 4245 f2fs_balance_fs(sbi, true); 4246 4247 file_start_write(filp); 4248 inode_lock(inode); 4249 4250 if (!f2fs_is_compress_backend_ready(inode)) { 4251 ret = -EOPNOTSUPP; 4252 goto out; 4253 } 4254 4255 if (!f2fs_compressed_file(inode) || 4256 is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) { 4257 ret = -EINVAL; 4258 goto out; 4259 } 4260 4261 ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX); 4262 if (ret) 4263 goto out; 4264 4265 set_inode_flag(inode, FI_ENABLE_COMPRESS); 4266 4267 last_idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE); 4268 4269 count = last_idx - page_idx; 4270 while (count && count >= cluster_size) { 4271 ret = redirty_blocks(inode, page_idx, cluster_size); 4272 if (ret < 0) 4273 break; 4274 4275 if (get_dirty_pages(inode) >= BLKS_PER_SEG(sbi)) { 4276 ret = filemap_fdatawrite(inode->i_mapping); 4277 if (ret < 0) 4278 break; 4279 } 4280 4281 count -= cluster_size; 4282 page_idx += cluster_size; 4283 4284 cond_resched(); 4285 if (fatal_signal_pending(current)) { 4286 ret = -EINTR; 4287 break; 4288 } 4289 } 4290 4291 if (!ret) 4292 ret = filemap_write_and_wait_range(inode->i_mapping, 0, 4293 LLONG_MAX); 4294 4295 clear_inode_flag(inode, FI_ENABLE_COMPRESS); 4296 4297 if (ret) 4298 f2fs_warn(sbi, "%s: The file might be partially compressed (errno=%d). Please delete the file.", 4299 __func__, ret); 4300 f2fs_update_time(sbi, REQ_TIME); 4301 out: 4302 inode_unlock(inode); 4303 file_end_write(filp); 4304 4305 return ret; 4306 } 4307 4308 static long __f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4309 { 4310 switch (cmd) { 4311 case FS_IOC_GETVERSION: 4312 return f2fs_ioc_getversion(filp, arg); 4313 case F2FS_IOC_START_ATOMIC_WRITE: 4314 return f2fs_ioc_start_atomic_write(filp, false); 4315 case F2FS_IOC_START_ATOMIC_REPLACE: 4316 return f2fs_ioc_start_atomic_write(filp, true); 4317 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 4318 return f2fs_ioc_commit_atomic_write(filp); 4319 case F2FS_IOC_ABORT_ATOMIC_WRITE: 4320 return f2fs_ioc_abort_atomic_write(filp); 4321 case F2FS_IOC_START_VOLATILE_WRITE: 4322 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 4323 return -EOPNOTSUPP; 4324 case F2FS_IOC_SHUTDOWN: 4325 return f2fs_ioc_shutdown(filp, arg); 4326 case FITRIM: 4327 return f2fs_ioc_fitrim(filp, arg); 4328 case FS_IOC_SET_ENCRYPTION_POLICY: 4329 return f2fs_ioc_set_encryption_policy(filp, arg); 4330 case FS_IOC_GET_ENCRYPTION_POLICY: 4331 return f2fs_ioc_get_encryption_policy(filp, arg); 4332 case FS_IOC_GET_ENCRYPTION_PWSALT: 4333 return f2fs_ioc_get_encryption_pwsalt(filp, arg); 4334 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 4335 return f2fs_ioc_get_encryption_policy_ex(filp, arg); 4336 case FS_IOC_ADD_ENCRYPTION_KEY: 4337 return f2fs_ioc_add_encryption_key(filp, arg); 4338 case FS_IOC_REMOVE_ENCRYPTION_KEY: 4339 return f2fs_ioc_remove_encryption_key(filp, arg); 4340 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 4341 return f2fs_ioc_remove_encryption_key_all_users(filp, arg); 4342 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 4343 return f2fs_ioc_get_encryption_key_status(filp, arg); 4344 case FS_IOC_GET_ENCRYPTION_NONCE: 4345 return f2fs_ioc_get_encryption_nonce(filp, arg); 4346 case F2FS_IOC_GARBAGE_COLLECT: 4347 return f2fs_ioc_gc(filp, arg); 4348 case F2FS_IOC_GARBAGE_COLLECT_RANGE: 4349 return f2fs_ioc_gc_range(filp, arg); 4350 case F2FS_IOC_WRITE_CHECKPOINT: 4351 return f2fs_ioc_write_checkpoint(filp); 4352 case F2FS_IOC_DEFRAGMENT: 4353 return f2fs_ioc_defragment(filp, arg); 4354 case F2FS_IOC_MOVE_RANGE: 4355 return f2fs_ioc_move_range(filp, arg); 4356 case F2FS_IOC_FLUSH_DEVICE: 4357 return f2fs_ioc_flush_device(filp, arg); 4358 case F2FS_IOC_GET_FEATURES: 4359 return f2fs_ioc_get_features(filp, arg); 4360 case F2FS_IOC_GET_PIN_FILE: 4361 return f2fs_ioc_get_pin_file(filp, arg); 4362 case F2FS_IOC_SET_PIN_FILE: 4363 return f2fs_ioc_set_pin_file(filp, arg); 4364 case F2FS_IOC_PRECACHE_EXTENTS: 4365 return f2fs_ioc_precache_extents(filp); 4366 case F2FS_IOC_RESIZE_FS: 4367 return f2fs_ioc_resize_fs(filp, arg); 4368 case FS_IOC_ENABLE_VERITY: 4369 return f2fs_ioc_enable_verity(filp, arg); 4370 case FS_IOC_MEASURE_VERITY: 4371 return f2fs_ioc_measure_verity(filp, arg); 4372 case FS_IOC_READ_VERITY_METADATA: 4373 return f2fs_ioc_read_verity_metadata(filp, arg); 4374 case FS_IOC_GETFSLABEL: 4375 return f2fs_ioc_getfslabel(filp, arg); 4376 case FS_IOC_SETFSLABEL: 4377 return f2fs_ioc_setfslabel(filp, arg); 4378 case F2FS_IOC_GET_COMPRESS_BLOCKS: 4379 return f2fs_ioc_get_compress_blocks(filp, arg); 4380 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 4381 return f2fs_release_compress_blocks(filp, arg); 4382 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 4383 return f2fs_reserve_compress_blocks(filp, arg); 4384 case F2FS_IOC_SEC_TRIM_FILE: 4385 return f2fs_sec_trim_file(filp, arg); 4386 case F2FS_IOC_GET_COMPRESS_OPTION: 4387 return f2fs_ioc_get_compress_option(filp, arg); 4388 case F2FS_IOC_SET_COMPRESS_OPTION: 4389 return f2fs_ioc_set_compress_option(filp, arg); 4390 case F2FS_IOC_DECOMPRESS_FILE: 4391 return f2fs_ioc_decompress_file(filp); 4392 case F2FS_IOC_COMPRESS_FILE: 4393 return f2fs_ioc_compress_file(filp); 4394 default: 4395 return -ENOTTY; 4396 } 4397 } 4398 4399 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 4400 { 4401 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp))))) 4402 return -EIO; 4403 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp)))) 4404 return -ENOSPC; 4405 4406 return __f2fs_ioctl(filp, cmd, arg); 4407 } 4408 4409 /* 4410 * Return %true if the given read or write request should use direct I/O, or 4411 * %false if it should use buffered I/O. 4412 */ 4413 static bool f2fs_should_use_dio(struct inode *inode, struct kiocb *iocb, 4414 struct iov_iter *iter) 4415 { 4416 unsigned int align; 4417 4418 if (!(iocb->ki_flags & IOCB_DIRECT)) 4419 return false; 4420 4421 if (f2fs_force_buffered_io(inode, iov_iter_rw(iter))) 4422 return false; 4423 4424 /* 4425 * Direct I/O not aligned to the disk's logical_block_size will be 4426 * attempted, but will fail with -EINVAL. 4427 * 4428 * f2fs additionally requires that direct I/O be aligned to the 4429 * filesystem block size, which is often a stricter requirement. 4430 * However, f2fs traditionally falls back to buffered I/O on requests 4431 * that are logical_block_size-aligned but not fs-block aligned. 4432 * 4433 * The below logic implements this behavior. 4434 */ 4435 align = iocb->ki_pos | iov_iter_alignment(iter); 4436 if (!IS_ALIGNED(align, i_blocksize(inode)) && 4437 IS_ALIGNED(align, bdev_logical_block_size(inode->i_sb->s_bdev))) 4438 return false; 4439 4440 return true; 4441 } 4442 4443 static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error, 4444 unsigned int flags) 4445 { 4446 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4447 4448 dec_page_count(sbi, F2FS_DIO_READ); 4449 if (error) 4450 return error; 4451 f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size); 4452 return 0; 4453 } 4454 4455 static const struct iomap_dio_ops f2fs_iomap_dio_read_ops = { 4456 .end_io = f2fs_dio_read_end_io, 4457 }; 4458 4459 static ssize_t f2fs_dio_read_iter(struct kiocb *iocb, struct iov_iter *to) 4460 { 4461 struct file *file = iocb->ki_filp; 4462 struct inode *inode = file_inode(file); 4463 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4464 struct f2fs_inode_info *fi = F2FS_I(inode); 4465 const loff_t pos = iocb->ki_pos; 4466 const size_t count = iov_iter_count(to); 4467 struct iomap_dio *dio; 4468 ssize_t ret; 4469 4470 if (count == 0) 4471 return 0; /* skip atime update */ 4472 4473 trace_f2fs_direct_IO_enter(inode, iocb, count, READ); 4474 4475 if (iocb->ki_flags & IOCB_NOWAIT) { 4476 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4477 ret = -EAGAIN; 4478 goto out; 4479 } 4480 } else { 4481 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4482 } 4483 4484 /* 4485 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4486 * the higher-level function iomap_dio_rw() in order to ensure that the 4487 * F2FS_DIO_READ counter will be decremented correctly in all cases. 4488 */ 4489 inc_page_count(sbi, F2FS_DIO_READ); 4490 dio = __iomap_dio_rw(iocb, to, &f2fs_iomap_ops, 4491 &f2fs_iomap_dio_read_ops, 0, NULL, 0); 4492 if (IS_ERR_OR_NULL(dio)) { 4493 ret = PTR_ERR_OR_ZERO(dio); 4494 if (ret != -EIOCBQUEUED) 4495 dec_page_count(sbi, F2FS_DIO_READ); 4496 } else { 4497 ret = iomap_dio_complete(dio); 4498 } 4499 4500 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4501 4502 file_accessed(file); 4503 out: 4504 trace_f2fs_direct_IO_exit(inode, pos, count, READ, ret); 4505 return ret; 4506 } 4507 4508 static void f2fs_trace_rw_file_path(struct file *file, loff_t pos, size_t count, 4509 int rw) 4510 { 4511 struct inode *inode = file_inode(file); 4512 char *buf, *path; 4513 4514 buf = f2fs_getname(F2FS_I_SB(inode)); 4515 if (!buf) 4516 return; 4517 path = dentry_path_raw(file_dentry(file), buf, PATH_MAX); 4518 if (IS_ERR(path)) 4519 goto free_buf; 4520 if (rw == WRITE) 4521 trace_f2fs_datawrite_start(inode, pos, count, 4522 current->pid, path, current->comm); 4523 else 4524 trace_f2fs_dataread_start(inode, pos, count, 4525 current->pid, path, current->comm); 4526 free_buf: 4527 f2fs_putname(buf); 4528 } 4529 4530 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to) 4531 { 4532 struct inode *inode = file_inode(iocb->ki_filp); 4533 const loff_t pos = iocb->ki_pos; 4534 ssize_t ret; 4535 4536 if (!f2fs_is_compress_backend_ready(inode)) 4537 return -EOPNOTSUPP; 4538 4539 if (trace_f2fs_dataread_start_enabled()) 4540 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4541 iov_iter_count(to), READ); 4542 4543 if (f2fs_should_use_dio(inode, iocb, to)) { 4544 ret = f2fs_dio_read_iter(iocb, to); 4545 } else { 4546 ret = filemap_read(iocb, to, 0); 4547 if (ret > 0) 4548 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4549 APP_BUFFERED_READ_IO, ret); 4550 } 4551 if (trace_f2fs_dataread_end_enabled()) 4552 trace_f2fs_dataread_end(inode, pos, ret); 4553 return ret; 4554 } 4555 4556 static ssize_t f2fs_file_splice_read(struct file *in, loff_t *ppos, 4557 struct pipe_inode_info *pipe, 4558 size_t len, unsigned int flags) 4559 { 4560 struct inode *inode = file_inode(in); 4561 const loff_t pos = *ppos; 4562 ssize_t ret; 4563 4564 if (!f2fs_is_compress_backend_ready(inode)) 4565 return -EOPNOTSUPP; 4566 4567 if (trace_f2fs_dataread_start_enabled()) 4568 f2fs_trace_rw_file_path(in, pos, len, READ); 4569 4570 ret = filemap_splice_read(in, ppos, pipe, len, flags); 4571 if (ret > 0) 4572 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4573 APP_BUFFERED_READ_IO, ret); 4574 4575 if (trace_f2fs_dataread_end_enabled()) 4576 trace_f2fs_dataread_end(inode, pos, ret); 4577 return ret; 4578 } 4579 4580 static ssize_t f2fs_write_checks(struct kiocb *iocb, struct iov_iter *from) 4581 { 4582 struct file *file = iocb->ki_filp; 4583 struct inode *inode = file_inode(file); 4584 ssize_t count; 4585 int err; 4586 4587 if (IS_IMMUTABLE(inode)) 4588 return -EPERM; 4589 4590 if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED)) 4591 return -EPERM; 4592 4593 count = generic_write_checks(iocb, from); 4594 if (count <= 0) 4595 return count; 4596 4597 err = file_modified(file); 4598 if (err) 4599 return err; 4600 return count; 4601 } 4602 4603 /* 4604 * Preallocate blocks for a write request, if it is possible and helpful to do 4605 * so. Returns a positive number if blocks may have been preallocated, 0 if no 4606 * blocks were preallocated, or a negative errno value if something went 4607 * seriously wrong. Also sets FI_PREALLOCATED_ALL on the inode if *all* the 4608 * requested blocks (not just some of them) have been allocated. 4609 */ 4610 static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter, 4611 bool dio) 4612 { 4613 struct inode *inode = file_inode(iocb->ki_filp); 4614 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4615 const loff_t pos = iocb->ki_pos; 4616 const size_t count = iov_iter_count(iter); 4617 struct f2fs_map_blocks map = {}; 4618 int flag; 4619 int ret; 4620 4621 /* If it will be an out-of-place direct write, don't bother. */ 4622 if (dio && f2fs_lfs_mode(sbi)) 4623 return 0; 4624 /* 4625 * Don't preallocate holes aligned to DIO_SKIP_HOLES which turns into 4626 * buffered IO, if DIO meets any holes. 4627 */ 4628 if (dio && i_size_read(inode) && 4629 (F2FS_BYTES_TO_BLK(pos) < F2FS_BLK_ALIGN(i_size_read(inode)))) 4630 return 0; 4631 4632 /* No-wait I/O can't allocate blocks. */ 4633 if (iocb->ki_flags & IOCB_NOWAIT) 4634 return 0; 4635 4636 /* If it will be a short write, don't bother. */ 4637 if (fault_in_iov_iter_readable(iter, count)) 4638 return 0; 4639 4640 if (f2fs_has_inline_data(inode)) { 4641 /* If the data will fit inline, don't bother. */ 4642 if (pos + count <= MAX_INLINE_DATA(inode)) 4643 return 0; 4644 ret = f2fs_convert_inline_inode(inode); 4645 if (ret) 4646 return ret; 4647 } 4648 4649 /* Do not preallocate blocks that will be written partially in 4KB. */ 4650 map.m_lblk = F2FS_BLK_ALIGN(pos); 4651 map.m_len = F2FS_BYTES_TO_BLK(pos + count); 4652 if (map.m_len > map.m_lblk) 4653 map.m_len -= map.m_lblk; 4654 else 4655 return 0; 4656 4657 map.m_may_create = true; 4658 if (dio) { 4659 map.m_seg_type = f2fs_rw_hint_to_seg_type(sbi, 4660 inode->i_write_hint); 4661 flag = F2FS_GET_BLOCK_PRE_DIO; 4662 } else { 4663 map.m_seg_type = NO_CHECK_TYPE; 4664 flag = F2FS_GET_BLOCK_PRE_AIO; 4665 } 4666 4667 ret = f2fs_map_blocks(inode, &map, flag); 4668 /* -ENOSPC|-EDQUOT are fine to report the number of allocated blocks. */ 4669 if (ret < 0 && !((ret == -ENOSPC || ret == -EDQUOT) && map.m_len > 0)) 4670 return ret; 4671 if (ret == 0) 4672 set_inode_flag(inode, FI_PREALLOCATED_ALL); 4673 return map.m_len; 4674 } 4675 4676 static ssize_t f2fs_buffered_write_iter(struct kiocb *iocb, 4677 struct iov_iter *from) 4678 { 4679 struct file *file = iocb->ki_filp; 4680 struct inode *inode = file_inode(file); 4681 ssize_t ret; 4682 4683 if (iocb->ki_flags & IOCB_NOWAIT) 4684 return -EOPNOTSUPP; 4685 4686 ret = generic_perform_write(iocb, from); 4687 4688 if (ret > 0) { 4689 f2fs_update_iostat(F2FS_I_SB(inode), inode, 4690 APP_BUFFERED_IO, ret); 4691 } 4692 return ret; 4693 } 4694 4695 static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error, 4696 unsigned int flags) 4697 { 4698 struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(iocb->ki_filp)); 4699 4700 dec_page_count(sbi, F2FS_DIO_WRITE); 4701 if (error) 4702 return error; 4703 f2fs_update_time(sbi, REQ_TIME); 4704 f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size); 4705 return 0; 4706 } 4707 4708 static void f2fs_dio_write_submit_io(const struct iomap_iter *iter, 4709 struct bio *bio, loff_t file_offset) 4710 { 4711 struct inode *inode = iter->inode; 4712 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4713 int seg_type = f2fs_rw_hint_to_seg_type(sbi, inode->i_write_hint); 4714 enum temp_type temp = f2fs_get_segment_temp(seg_type); 4715 4716 bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi, DATA, temp); 4717 submit_bio(bio); 4718 } 4719 4720 static const struct iomap_dio_ops f2fs_iomap_dio_write_ops = { 4721 .end_io = f2fs_dio_write_end_io, 4722 .submit_io = f2fs_dio_write_submit_io, 4723 }; 4724 4725 static void f2fs_flush_buffered_write(struct address_space *mapping, 4726 loff_t start_pos, loff_t end_pos) 4727 { 4728 int ret; 4729 4730 ret = filemap_write_and_wait_range(mapping, start_pos, end_pos); 4731 if (ret < 0) 4732 return; 4733 invalidate_mapping_pages(mapping, 4734 start_pos >> PAGE_SHIFT, 4735 end_pos >> PAGE_SHIFT); 4736 } 4737 4738 static ssize_t f2fs_dio_write_iter(struct kiocb *iocb, struct iov_iter *from, 4739 bool *may_need_sync) 4740 { 4741 struct file *file = iocb->ki_filp; 4742 struct inode *inode = file_inode(file); 4743 struct f2fs_inode_info *fi = F2FS_I(inode); 4744 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 4745 const bool do_opu = f2fs_lfs_mode(sbi); 4746 const loff_t pos = iocb->ki_pos; 4747 const ssize_t count = iov_iter_count(from); 4748 unsigned int dio_flags; 4749 struct iomap_dio *dio; 4750 ssize_t ret; 4751 4752 trace_f2fs_direct_IO_enter(inode, iocb, count, WRITE); 4753 4754 if (iocb->ki_flags & IOCB_NOWAIT) { 4755 /* f2fs_convert_inline_inode() and block allocation can block */ 4756 if (f2fs_has_inline_data(inode) || 4757 !f2fs_overwrite_io(inode, pos, count)) { 4758 ret = -EAGAIN; 4759 goto out; 4760 } 4761 4762 if (!f2fs_down_read_trylock(&fi->i_gc_rwsem[WRITE])) { 4763 ret = -EAGAIN; 4764 goto out; 4765 } 4766 if (do_opu && !f2fs_down_read_trylock(&fi->i_gc_rwsem[READ])) { 4767 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4768 ret = -EAGAIN; 4769 goto out; 4770 } 4771 } else { 4772 ret = f2fs_convert_inline_inode(inode); 4773 if (ret) 4774 goto out; 4775 4776 f2fs_down_read(&fi->i_gc_rwsem[WRITE]); 4777 if (do_opu) 4778 f2fs_down_read(&fi->i_gc_rwsem[READ]); 4779 } 4780 4781 /* 4782 * We have to use __iomap_dio_rw() and iomap_dio_complete() instead of 4783 * the higher-level function iomap_dio_rw() in order to ensure that the 4784 * F2FS_DIO_WRITE counter will be decremented correctly in all cases. 4785 */ 4786 inc_page_count(sbi, F2FS_DIO_WRITE); 4787 dio_flags = 0; 4788 if (pos + count > inode->i_size) 4789 dio_flags |= IOMAP_DIO_FORCE_WAIT; 4790 dio = __iomap_dio_rw(iocb, from, &f2fs_iomap_ops, 4791 &f2fs_iomap_dio_write_ops, dio_flags, NULL, 0); 4792 if (IS_ERR_OR_NULL(dio)) { 4793 ret = PTR_ERR_OR_ZERO(dio); 4794 if (ret == -ENOTBLK) 4795 ret = 0; 4796 if (ret != -EIOCBQUEUED) 4797 dec_page_count(sbi, F2FS_DIO_WRITE); 4798 } else { 4799 ret = iomap_dio_complete(dio); 4800 } 4801 4802 if (do_opu) 4803 f2fs_up_read(&fi->i_gc_rwsem[READ]); 4804 f2fs_up_read(&fi->i_gc_rwsem[WRITE]); 4805 4806 if (ret < 0) 4807 goto out; 4808 if (pos + ret > inode->i_size) 4809 f2fs_i_size_write(inode, pos + ret); 4810 if (!do_opu) 4811 set_inode_flag(inode, FI_UPDATE_WRITE); 4812 4813 if (iov_iter_count(from)) { 4814 ssize_t ret2; 4815 loff_t bufio_start_pos = iocb->ki_pos; 4816 4817 /* 4818 * The direct write was partial, so we need to fall back to a 4819 * buffered write for the remainder. 4820 */ 4821 4822 ret2 = f2fs_buffered_write_iter(iocb, from); 4823 if (iov_iter_count(from)) 4824 f2fs_write_failed(inode, iocb->ki_pos); 4825 if (ret2 < 0) 4826 goto out; 4827 4828 /* 4829 * Ensure that the pagecache pages are written to disk and 4830 * invalidated to preserve the expected O_DIRECT semantics. 4831 */ 4832 if (ret2 > 0) { 4833 loff_t bufio_end_pos = bufio_start_pos + ret2 - 1; 4834 4835 ret += ret2; 4836 4837 f2fs_flush_buffered_write(file->f_mapping, 4838 bufio_start_pos, 4839 bufio_end_pos); 4840 } 4841 } else { 4842 /* iomap_dio_rw() already handled the generic_write_sync(). */ 4843 *may_need_sync = false; 4844 } 4845 out: 4846 trace_f2fs_direct_IO_exit(inode, pos, count, WRITE, ret); 4847 return ret; 4848 } 4849 4850 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from) 4851 { 4852 struct inode *inode = file_inode(iocb->ki_filp); 4853 const loff_t orig_pos = iocb->ki_pos; 4854 const size_t orig_count = iov_iter_count(from); 4855 loff_t target_size; 4856 bool dio; 4857 bool may_need_sync = true; 4858 int preallocated; 4859 const loff_t pos = iocb->ki_pos; 4860 const ssize_t count = iov_iter_count(from); 4861 ssize_t ret; 4862 4863 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) { 4864 ret = -EIO; 4865 goto out; 4866 } 4867 4868 if (!f2fs_is_compress_backend_ready(inode)) { 4869 ret = -EOPNOTSUPP; 4870 goto out; 4871 } 4872 4873 if (iocb->ki_flags & IOCB_NOWAIT) { 4874 if (!inode_trylock(inode)) { 4875 ret = -EAGAIN; 4876 goto out; 4877 } 4878 } else { 4879 inode_lock(inode); 4880 } 4881 4882 if (f2fs_is_pinned_file(inode) && 4883 !f2fs_overwrite_io(inode, pos, count)) { 4884 ret = -EIO; 4885 goto out_unlock; 4886 } 4887 4888 ret = f2fs_write_checks(iocb, from); 4889 if (ret <= 0) 4890 goto out_unlock; 4891 4892 /* Determine whether we will do a direct write or a buffered write. */ 4893 dio = f2fs_should_use_dio(inode, iocb, from); 4894 4895 /* Possibly preallocate the blocks for the write. */ 4896 target_size = iocb->ki_pos + iov_iter_count(from); 4897 preallocated = f2fs_preallocate_blocks(iocb, from, dio); 4898 if (preallocated < 0) { 4899 ret = preallocated; 4900 } else { 4901 if (trace_f2fs_datawrite_start_enabled()) 4902 f2fs_trace_rw_file_path(iocb->ki_filp, iocb->ki_pos, 4903 orig_count, WRITE); 4904 4905 /* Do the actual write. */ 4906 ret = dio ? 4907 f2fs_dio_write_iter(iocb, from, &may_need_sync) : 4908 f2fs_buffered_write_iter(iocb, from); 4909 4910 if (trace_f2fs_datawrite_end_enabled()) 4911 trace_f2fs_datawrite_end(inode, orig_pos, ret); 4912 } 4913 4914 /* Don't leave any preallocated blocks around past i_size. */ 4915 if (preallocated && i_size_read(inode) < target_size) { 4916 f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4917 filemap_invalidate_lock(inode->i_mapping); 4918 if (!f2fs_truncate(inode)) 4919 file_dont_truncate(inode); 4920 filemap_invalidate_unlock(inode->i_mapping); 4921 f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]); 4922 } else { 4923 file_dont_truncate(inode); 4924 } 4925 4926 clear_inode_flag(inode, FI_PREALLOCATED_ALL); 4927 out_unlock: 4928 inode_unlock(inode); 4929 out: 4930 trace_f2fs_file_write_iter(inode, orig_pos, orig_count, ret); 4931 4932 if (ret > 0 && may_need_sync) 4933 ret = generic_write_sync(iocb, ret); 4934 4935 /* If buffered IO was forced, flush and drop the data from 4936 * the page cache to preserve O_DIRECT semantics 4937 */ 4938 if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT)) 4939 f2fs_flush_buffered_write(iocb->ki_filp->f_mapping, 4940 orig_pos, 4941 orig_pos + ret - 1); 4942 4943 return ret; 4944 } 4945 4946 static int f2fs_file_fadvise(struct file *filp, loff_t offset, loff_t len, 4947 int advice) 4948 { 4949 struct address_space *mapping; 4950 struct backing_dev_info *bdi; 4951 struct inode *inode = file_inode(filp); 4952 int err; 4953 4954 if (advice == POSIX_FADV_SEQUENTIAL) { 4955 if (S_ISFIFO(inode->i_mode)) 4956 return -ESPIPE; 4957 4958 mapping = filp->f_mapping; 4959 if (!mapping || len < 0) 4960 return -EINVAL; 4961 4962 bdi = inode_to_bdi(mapping->host); 4963 filp->f_ra.ra_pages = bdi->ra_pages * 4964 F2FS_I_SB(inode)->seq_file_ra_mul; 4965 spin_lock(&filp->f_lock); 4966 filp->f_mode &= ~FMODE_RANDOM; 4967 spin_unlock(&filp->f_lock); 4968 return 0; 4969 } else if (advice == POSIX_FADV_WILLNEED && offset == 0) { 4970 /* Load extent cache at the first readahead. */ 4971 f2fs_precache_extents(inode); 4972 } 4973 4974 err = generic_fadvise(filp, offset, len, advice); 4975 if (!err && advice == POSIX_FADV_DONTNEED && 4976 test_opt(F2FS_I_SB(inode), COMPRESS_CACHE) && 4977 f2fs_compressed_file(inode)) 4978 f2fs_invalidate_compress_pages(F2FS_I_SB(inode), inode->i_ino); 4979 4980 return err; 4981 } 4982 4983 #ifdef CONFIG_COMPAT 4984 struct compat_f2fs_gc_range { 4985 u32 sync; 4986 compat_u64 start; 4987 compat_u64 len; 4988 }; 4989 #define F2FS_IOC32_GARBAGE_COLLECT_RANGE _IOW(F2FS_IOCTL_MAGIC, 11,\ 4990 struct compat_f2fs_gc_range) 4991 4992 static int f2fs_compat_ioc_gc_range(struct file *file, unsigned long arg) 4993 { 4994 struct compat_f2fs_gc_range __user *urange; 4995 struct f2fs_gc_range range; 4996 int err; 4997 4998 urange = compat_ptr(arg); 4999 err = get_user(range.sync, &urange->sync); 5000 err |= get_user(range.start, &urange->start); 5001 err |= get_user(range.len, &urange->len); 5002 if (err) 5003 return -EFAULT; 5004 5005 return __f2fs_ioc_gc_range(file, &range); 5006 } 5007 5008 struct compat_f2fs_move_range { 5009 u32 dst_fd; 5010 compat_u64 pos_in; 5011 compat_u64 pos_out; 5012 compat_u64 len; 5013 }; 5014 #define F2FS_IOC32_MOVE_RANGE _IOWR(F2FS_IOCTL_MAGIC, 9, \ 5015 struct compat_f2fs_move_range) 5016 5017 static int f2fs_compat_ioc_move_range(struct file *file, unsigned long arg) 5018 { 5019 struct compat_f2fs_move_range __user *urange; 5020 struct f2fs_move_range range; 5021 int err; 5022 5023 urange = compat_ptr(arg); 5024 err = get_user(range.dst_fd, &urange->dst_fd); 5025 err |= get_user(range.pos_in, &urange->pos_in); 5026 err |= get_user(range.pos_out, &urange->pos_out); 5027 err |= get_user(range.len, &urange->len); 5028 if (err) 5029 return -EFAULT; 5030 5031 return __f2fs_ioc_move_range(file, &range); 5032 } 5033 5034 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 5035 { 5036 if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file))))) 5037 return -EIO; 5038 if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(file)))) 5039 return -ENOSPC; 5040 5041 switch (cmd) { 5042 case FS_IOC32_GETVERSION: 5043 cmd = FS_IOC_GETVERSION; 5044 break; 5045 case F2FS_IOC32_GARBAGE_COLLECT_RANGE: 5046 return f2fs_compat_ioc_gc_range(file, arg); 5047 case F2FS_IOC32_MOVE_RANGE: 5048 return f2fs_compat_ioc_move_range(file, arg); 5049 case F2FS_IOC_START_ATOMIC_WRITE: 5050 case F2FS_IOC_START_ATOMIC_REPLACE: 5051 case F2FS_IOC_COMMIT_ATOMIC_WRITE: 5052 case F2FS_IOC_START_VOLATILE_WRITE: 5053 case F2FS_IOC_RELEASE_VOLATILE_WRITE: 5054 case F2FS_IOC_ABORT_ATOMIC_WRITE: 5055 case F2FS_IOC_SHUTDOWN: 5056 case FITRIM: 5057 case FS_IOC_SET_ENCRYPTION_POLICY: 5058 case FS_IOC_GET_ENCRYPTION_PWSALT: 5059 case FS_IOC_GET_ENCRYPTION_POLICY: 5060 case FS_IOC_GET_ENCRYPTION_POLICY_EX: 5061 case FS_IOC_ADD_ENCRYPTION_KEY: 5062 case FS_IOC_REMOVE_ENCRYPTION_KEY: 5063 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS: 5064 case FS_IOC_GET_ENCRYPTION_KEY_STATUS: 5065 case FS_IOC_GET_ENCRYPTION_NONCE: 5066 case F2FS_IOC_GARBAGE_COLLECT: 5067 case F2FS_IOC_WRITE_CHECKPOINT: 5068 case F2FS_IOC_DEFRAGMENT: 5069 case F2FS_IOC_FLUSH_DEVICE: 5070 case F2FS_IOC_GET_FEATURES: 5071 case F2FS_IOC_GET_PIN_FILE: 5072 case F2FS_IOC_SET_PIN_FILE: 5073 case F2FS_IOC_PRECACHE_EXTENTS: 5074 case F2FS_IOC_RESIZE_FS: 5075 case FS_IOC_ENABLE_VERITY: 5076 case FS_IOC_MEASURE_VERITY: 5077 case FS_IOC_READ_VERITY_METADATA: 5078 case FS_IOC_GETFSLABEL: 5079 case FS_IOC_SETFSLABEL: 5080 case F2FS_IOC_GET_COMPRESS_BLOCKS: 5081 case F2FS_IOC_RELEASE_COMPRESS_BLOCKS: 5082 case F2FS_IOC_RESERVE_COMPRESS_BLOCKS: 5083 case F2FS_IOC_SEC_TRIM_FILE: 5084 case F2FS_IOC_GET_COMPRESS_OPTION: 5085 case F2FS_IOC_SET_COMPRESS_OPTION: 5086 case F2FS_IOC_DECOMPRESS_FILE: 5087 case F2FS_IOC_COMPRESS_FILE: 5088 break; 5089 default: 5090 return -ENOIOCTLCMD; 5091 } 5092 return __f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5093 } 5094 #endif 5095 5096 const struct file_operations f2fs_file_operations = { 5097 .llseek = f2fs_llseek, 5098 .read_iter = f2fs_file_read_iter, 5099 .write_iter = f2fs_file_write_iter, 5100 .iopoll = iocb_bio_iopoll, 5101 .open = f2fs_file_open, 5102 .release = f2fs_release_file, 5103 .mmap = f2fs_file_mmap, 5104 .flush = f2fs_file_flush, 5105 .fsync = f2fs_sync_file, 5106 .fallocate = f2fs_fallocate, 5107 .unlocked_ioctl = f2fs_ioctl, 5108 #ifdef CONFIG_COMPAT 5109 .compat_ioctl = f2fs_compat_ioctl, 5110 #endif 5111 .splice_read = f2fs_file_splice_read, 5112 .splice_write = iter_file_splice_write, 5113 .fadvise = f2fs_file_fadvise, 5114 .fop_flags = FOP_BUFFER_RASYNC, 5115 }; 5116